Nucl - Phys.B v.787

Nuclear Physics B 787 (2007) 154
Squark and gaugino hadroproduction and decays

in non-minimal flavour violating supersymmetry
Giuseppe Bozzi a , Benjamin Fuks b , Bjrn Herrmann b , Michael Klasen b,
a Institut fr Theoretische Physik, Universitt Karlsruhe, Postfach 6980, D-76128 Karlsruhe, Germany
b Laboratoire de Physique Subatomique et de Cosmologie, Universit Joseph Fourier/CNRS-IN2P3,
53 Avenue des Martyrs, F-38026 Grenoble, France

Received 11 April 2007; received in revised form 28 May 2007; accepted 29 May 2007
Available online 14 June 2007
Abstract
We present an extensive analysis of squark and gaugino hadroproduction and decays in non-minimal
flavour violating supersymmetry. We employ the so-called super-CKM basis to define the possible misalignment of quark and squark rotations, and we use generalized (possibly complex) charges to define the mutual
couplings of (s)quarks and gauge bosons/gauginos. The cross sections for all squark(anti-)squark/gaugino
pair and squark-gaugino associated production processes as well as their decay widths are then given in
compact analytic form. For four different constrained supersymmetry breaking models with non-minimal
flavour violation in the second/third generation squark sector only, we establish the parameter space regions allowed/favoured by low-energy, electroweak precision, and cosmological constraints and display the
chirality and flavour decomposition of all up- and down-type squark mass eigenstates. Finally, we compute numerically the dependence of a representative sample of production cross sections at the LHC on the
off-diagonal mass matrix elements in the experimentally allowed/favoured ranges.
2007 Elsevier B.V. All rights reserved.
1. Introduction
Weak scale supersymmetry (SUSY) remains a both theoretically and phenomenologically
attractive extension of the Standard Model (SM) of particle physics [1,2]. Apart from linking
bosons with fermions and unifying internal and external (spacetime) symmetries, SUSY allows
for a stabilization of the gap between the Planck and the electroweak scale and for gauge coupling
* Corresponding author.
E-mail address: klasen@lpsc.in2p3.fr (M. Klasen).

0550-3213/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2007.05.031
G. Bozzi et al. / Nuclear Physics B 787 (2007) 154
unification at high energies. It appears naturally in string theories, includes gravity, and contains a
stable lightest SUSY particle (LSP) as a dark matter candidate. Spin partners of the SM particles
have not yet been observed, and in order to remain a viable solution to the hierarchy problem,
SUSY must be broken at low energy via soft mass terms in the Lagrangian. As a consequence,
the SUSY particles must be massive in comparison to their SM counterparts, and the Tevatron
and the LHC will perform a conclusive search covering a wide range of masses up to the TeV
scale.
If SUSY particles exist, they should also appear in virtual particle loops and affect low-energy
and electroweak precision observables. In particular, flavour-changing neutral currents (FCNC),
which appear only at the one-loop level even in the SM, put severe constraints on new physics
contributions appearing at the same perturbative order. Extended technicolour and many composite models have thus been ruled out, while the Minimal Supersymmetric Standard Model
(MSSM) has passed these crucial tests. This is largely due to the assumption of constrained Minimal Flavour Violation (cMFV) [3,4] or Minimal Flavour Violation (MFV) [57], where heavy
SUSY particles may appear in the loops, but flavour changes are either neglected or completely
dictated by the structure of the Yukawa couplings and thus the CKM-matrix [8,9].
The squark mass matrices M 2 , M 2 , and M 2 are usually expressed in the super-CKM flavour
Q
U
D
basis [10]. In MFV SUSY scenarios, their flavour violating non-diagonal entries ij , where
i, j = L, R refer to the helicity of the (SM partner of the) squark, stem from the trilinear Yukawa
couplings of the fermion and Higgs supermultiplets and the resulting different renormalizations
of the quark and squark mass matrices, which induce additional flavour violation at the weak
scale through renormalization group running [1114], while in cMFV scenarios, these flavour
violating off-diagonal entries are simply neglected at both the SUSY-breaking and the weak
scale.
When SUSY is embedded in larger structures such as grand unified theories (GUTs), new
sources of flavour violation can appear [15]. For example, local gauge symmetry allows for Rparity violating terms in the SUSY Lagrangian, but these terms are today severely constrained by
proton decay and collider searches. In non-minimal flavour violating (NMFV) SUSY, additional
sources of flavour violation are included in the mass matrices at the weak scale, and their flavour
violating off-diagonal terms cannot be simply deduced from the CKM matrix alone. NMFV is
then conveniently parameterized in the super-CKM basis by considering them as free parameters.
The scaling of these entries with the SUSY-breaking scale MSUSY implies a hierarchy LL
LR,RL RR [15].
Squark mixing is expected to be largest for the second and third generations due to the large
Yukawa couplings involved [16]. In addition, stringent experimental constraints for the first generation are imposed by precise measurements of K 0 K 0 mixing and first evidence of D 0 D 0
mixing [1719]. Furthermore, direct searches of flavour violation depend on the possibility of
flavour tagging, established experimentally only for heavy flavours. We therefore consider here
only flavour mixings of second- and third-generation squarks.
The direct search for SUSY particles constitutes a major physics goal of present (Tevatron)
and future (LHC) hadron colliders. SUSY particle hadroproduction and decay has therefore been
studied in detail theoretically. Next-to-leading order (NLO) SUSY-QCD calculations exist for the
production of squarks and gluinos [20], sleptons [21], and gauginos [22] as well as for their associated production [23]. The production of top [24] and bottom [25] squarks with large helicity
mixing has received particular attention. Recently, both QCD one-loop and electroweak treelevel contributions have been calculated for non-diagonal, diagonal and mixed top and bottom
squark pair production [26]. However, flavour violation has never been considered in the con-
text of collider searches apart from the CKM-matrix appearing in the electroweak stopsbottom
production channel [26].
It is the aim of this paper to investigate for the first time the possible effects of non-minimal
flavour violation at hadron colliders. To this end, we re-calculate all squark and gaugino production and decay helicity amplitudes, keeping at the same time the CKM-matrix and the quark
masses to account for non-diagonal charged-current gaugino and higgsino Yukawa interactions,
and generalizing the two-dimensional helicity mixing matrices, often assumed to be real, to generally complex six-dimensional helicity and generational mixing matrices. We keep the notation
compact by presenting all analytical expressions in terms of generalized couplings. In order
to obtain numerical predictions for hadron colliders, we have implemented all our results in a
flexible computer programme. In our phenomenological analysis of NMFV squark and gaugino
production, we concentrate on the LHC due to its larger centre-of-mass energy and luminosity.
We pay particular attention to the interesting interplay of parton density functions (PDFs), which
are dominated by light quarks, strong gluino contributions, which are generally larger than electroweak contributions and need not be flavour-diagonal, and the appearance of third-generation
squarks in the final state, which are easily identified experimentally and generally lighter than
first- and second-generation squarks.
After reviewing the MSSM with NMFV and setting up our notation in Section 2, we define
in Section 3 generalized couplings of quarks, squarks, gauge bosons, and gauginos. We then use
these couplings to present our analytical calculations in concise form. In particular, we have computed partonic cross sections for NMFV squarkantisquark and squarksquark pair production,
squark and gaugino associated and gaugino pair production as well as NMFV two-body decay
widths of all squarks and gauginos. Section 4 is devoted to a precise numerical analysis of the
experimentally allowed NMFV SUSY parameter space, an investigation of the corresponding
helicity and flavour decomposition of the up- and down-type squarks, and the definition of four
collider-friendly benchmark points. These points are then investigated in detail in Section 5 so as
to determine the possible sensitivity of the LHC experiments on the allowed NMFV parameter
regions in the above-mentioned production channels. Our conclusions are presented in Section 6.
2. Non-minimal flavour violation in the MSSM
Within the SM, the only source of flavour violation arises through the rotation of the up-type
) to the basis of physical mass eigenstates
(down-type) quark interaction eigenstates uL,R (dL,R
uL,R (dL,R ), such that
d

dL,R = VL,R
dL,R
u
and uL,R = VL,R
uL,R .
(1)
u,d
The four bi-unitary matrices VL,R
diagonalize the quark Yukawa matrices and render the
charged-current interactions proportional to the unitary CKM-matrix [8,9]

Vud Vus Vub
u d
V = VL VL = Vcd Vcs Vcb .
(2)
Vtd Vts Vtb
In the super-CKM basis [10], the squark interaction eigenstates undergo the same rotations at
high energy scale as their quark counterparts, so that their charged-current interactions are also
proportional to the SM CKM-matrix. However, different renormalizations of quarks and squarks
introduce a mismatch of quark and squark field rotations at low energies, so that the squark mass
matrices effectively become non-diagonal [1114]. NMFV is then conveniently parameterized

qq
by (arbitrary) non-diagonal entries ij with i, j = L, R in the squared squark mass matrices
M 2
Lu
cu
LL
tu
LL
2
Mu =
mu Xu
cu
LR
tu
LR
ut
uc
ut
uc
LL LL mu Xu LR LR
cu
ct
ct
LL RL mc Xc LR
tc
M 2 tu
RL RL mt Xt
Lt
uc
ut
2
ut
RL MR u RR RR
ct
cu
ct
2
mc Xc RL RR M RR
Rc
tc
tu
tc
2
LR mt Xt RR RR M
M 2
Lc
tc
LL
uc
RL
(3)
Rt
and
M 2
Ld
sd
LL
bd
LL
Md2 =
m X
d d
sd
LR
bd
LR
db
ds
db
ds
LL LL md Xd LR LR
M 2
Ls
bs
LL
ds
RL
sb
LL
M 2
Lb
db
RL
ms Xs sb
RL
bs
LR mb Xb
sb
sd
RL ms Xs LR
bd
bs
RL RL mb Xb
,
ds
db
2
M RR RR
Rd
sd
sb
2
RR M RR
Rs
bs
2
bd
M
RR
RR
(4)
Rb
where the diagonal elements are given by
3

2
2
2
2
ML2 = MQ
+ mq + cos 2MZ Tq eq sW ,
q
(5)
2
MR2 = MU2 + m2q + cos 2MZ2 eq sW
for up-type squarks,
(6)
2
2
2
2
MR2 = MD
+ mq + cos 2MZ eq sW
for down-type squarks,
(7)
while the well-known squark helicity mixing is generated by the elements

cot for up-type squarks,
Xq = Aq
tan for down-type squarks.
(8)
Here, mq , Tq3 , and eq denote the mass, weak isospin quantum number, and electric charge of the
quark q. mZ is the Z-boson mass, and sW (cW ) is the sine (cosine) of the electroweak mixing angle W . The soft SUSY-breaking mass terms are MQ q and MU q ,D q for the left- and right-handed
squarks. Aq and are the trilinear coupling and off-diagonal Higgs mass parameter, respectively,
and tan = vu /vd is the ratio of vacuum expectation values of the two Higgs doublets. The scalqq
ing of the flavour violating entries ij with the SUSY-breaking scale MSUSY implies a hierarchy
qq
qq
qq
LL LR,RL RR among them [15]. They are usually normalized to the diagonal entries
[18], so that
qq
qq
ij = ij Miq Mj .
(9)
qq
Note also that SU(2) gauge invariance relates the (numerically largest) LL of up- and downtype quarks through the CKM-matrix, implying that a large difference between them is not
allowed.
The diagonalization of the mass matrices Mu2 and M 2 requires the introduction of two addi-
tional 6 6 matrices R u and R d with

diag m2u 1 , . . . , m2u 6 = R u Mu2 R u

diag m2d , . . . , m2d = R d Md2 R d .
and
(10)
By convention, the masses are ordered according to mq1 < < mq6 . The physical mass eigenstates are given by

d1
u L
dL
u 1
d2
cL
sL
u 2

d
t
b
u 3
(11)
= R u L and 3 = R d L .
d4
u R
dR
u 4

u 5
cR
sR
d5
u 6
d6
bR
tR
qq
In the limit of vanishing off-diagonal parameters ij , the matrices R q become flavour-diagonal,

leaving only the well-known helicity mixing already present in cMFV.
3. Analytical results for production cross sections and decay widths
In this section, we introduce concise definitions of generalized strong and electroweak couplings in NMFV SUSY and compute analytically the corresponding partonic cross sections for
squark and gaugino production as well as their decay widths. The cross sections of the production
processes
()
()
qi (p1 )qj (p2 ),

aha (pa )bhb (pb )
j (p1 )qi (p2 ),
(0) (p ) (0) (p )
1 j
2
i
()
(12)
g and expressed in terms

are presented for definite helicities ha,b of the initial partons a, b = q, q,
of the squark, chargino, neutralino, and gluino masses mqk , m , m 0 , and mg , the conventional
k
k
Mandelstam variables,
s = (pa + pb )2 ,
t = (pa p1 )2 ,
and u = (pa p2 )2 ,
(13)
and the masses of the neutral and charged electroweak gauge bosons mZ and mW . Propagators
appear as mass-subtracted Mandelstam variables,
sw = s m2W ,
sz = s m2Z ,
t 0 = t m2 0 ,
u 0 = u m2 0 ,
m2 ,
j
uj = u m2 ,
j
tj = t
tg = t m2g ,
ug = u m2g ,
tqi = t m2qi ,
uqi = u m2qi .
(14)
Unpolarized cross sections, averaged over initial spins, can easily be derived from the expression
d =
d 1,1 + d 1,1 + d 1,1 + d 1,1

,
4
(15)
while single- and double-polarized cross sections, including the same average factor for initial
spins, are given by
d 1,1 + d 1,1 d 1,1 d 1,1
and
4
d 1,1 d 1,1 d 1,1 + d 1,1
d LL =
,
4
so that the single- and double-spin asymmetries become
d L =
AL =
d L
d
and ALL =
d LL
.
d
(16)
(17)
3.1. Generalized strong and electroweak couplings in NMFV SUSY

Considering the strong interaction first, it is well known that
the interaction of quarks, squarks,
and gluinos, whose coupling is normally just given by gs = 4s , can in general lead to flavour
violation in the left- and right-handed sectors through non-diagonal entries in the matrices R q ,

q
q
{Lqj qk g , Rqj qk g } = Rj k , Rj (k+3) .
(18)
Of course, the involved quark and squark both have to be up- or down-type, since the gluino is
electrically neutral.
2 =
For the electroweak interaction, we define the square of the weak coupling constant gW
2
2
2
e / sin W in terms of the electromagnetic fine structure constant = e /(4) and the squared
2 = 1 cos2 = 1 c2 . Following
sine of the electroweak mixing angle xW = sin2 W = sW
W
W
the standard notation, the W i j , Z i j , and Z i0 j0 interaction vertices

are proportional to [2]
1
1
OijL = Ni4 Vj2 + Ni2 Vj1 and OijR = Ni3

Uj 2 + Ni2
Uj 1 ,
2
2
1
1
Oij L = Vi1 Vj1 Vi2 Vj2 + ij xW and Oij R = Ui1

Uj 1 Ui2
Uj 2 + ij xW ,
2
2
1
1
1
1
Nj 3 Ni4
Nj 4 .
Oij L = Ni3 Nj3 + Ni4 Nj4 and Oij R = Ni3
(19)
2
2
2
2
In NMFV, the coupling strengths of left- and right-handed (s)quarks to the electroweak gauge
bosons are given by

{Lqq Z , Rqq Z } = 2Tq3 2eq xW qq ,

3

u u u

{Lqi qj Z , Rqi qj Z } = 2Tq3 2eq xW

Rik Rj k , Ri(3+k) Rju(3+k) ,
{Lqq W , Rqq W } =
{Lu i dj W , Ru i dj W } =

2cW Vqq , 0 ,
k=1
3

u d
2cW Vuk dl Rik
Rj l , 0 ,
(20)
k,l=1
3
3
where the weak isospin quantum numbers are T{q,
q}
= 1/2 for left-handed and T{q,q}
= 0 for
right-handed up- and down-type (s)quarks, their fractional electromagnetic charges are denoted
by e{q,q}
, and Vkl are the elements of the CKM-matrix defined in Eq. (2). To simplify the notation,
we have introduced flavour indices in the latter, d1 = d, d2 = s, d3 = b, u1 = u, u2 = c, and
u3 = t .
The SUSY counterparts of these vertices correspond to the quarksquarkgaugino couplings,

Ldj dk 0 =
i
Rd
0
j dk i
mdk cW Ni3 Rjd(k+3)

eq Tq3 sW Ni1 + Tq3 cW Ni2 Rjdk +
,
2mW cos
= eq sW Ni1 Rjd(k+3)
mdk cW Ni3 Rjdk

2mW cos
muk cW Ni4 Rju(k+3)

eq Tq3 sW Ni1 + Tq3 cW Ni2 Rjuk +
,
i
2mW sin
muk cW Ni4 Rjuk
u
,
Ru u 0 = eq sW Ni1 Rj (k+3)
j k i
2mW sin

3

mdk Ui2 Rjd(k+3)
d
Ldj ul =
Ui1 Rj k
Vul dk ,
i
2mW cos
k=1

Lu j uk 0 =
Rd
j ul i
Lu j dl
i
Ru
3 m V V Rd

ul i2 ul dk j k
,
2m
sin
W
k=1

3

muk Vi2 Rju(k+3)
u
=
Vi1 Rj k
Vuk dl ,
2mW sin
k=1
j dl i
3 m U V R u

dl i2 uk dl j k
,
2mW cos
k=1
(21)
where the matrices N , U and V relate to the gaugino/higgsino mixing (see Appendix A). All
other couplings vanish due to (electromagnetic) charge conservation (e.g. Lu j ul ). These geni
eral expressions can be simplified by neglecting the Yukawa couplings except for the one of the
top quark, whose mass is not small compared to mW . For the sake of simplicity, we use the
generic notation
1

2
Cabc , Cabc
= {Labc , Rabc }
(22)
in the following.
3.2. Squarkantisquark pair production
The production of charged squarkantisquark pairs
q(ha , pa )q (hb , pb ) u i (p1 )dj (p2 ),
(23)
where i, j = 1, . . . , 6 label up- and down-type squark mass eigenstates, ha,b helicities, and
pa,b,1,2 four-momenta, proceeds from an equally charged quarkantiquark initial state through
the tree-level Feynman diagrams shown in Fig. 1. The corresponding cross section can be written
Fig. 1. Tree-level Feynman diagrams for the production of charged squarkantisquark pairs in quarkantiquark collisions.
in a compact way as
q q
d ha ,hb
dt

kl

N11
W
G11
= (1 ha )(1 + hb ) 2 +
+ 2
t 0t 0
sw
tg
k,l=1,...,4 k l

[N W]k
[GW]
+
+
t 0 sw
tg sw
k=1,...,4 k

kl

N12
G12
+ (1 ha )(1 hb )
+ 2
t 0t 0
tg
k,l=1,...,4 k l

kl

N21
G21
+ (1 + ha )(1 + hb )
+ 2
t 0t 0
tg
k,l=1,...,4 k l

kl

N22
G22
+ (1 + ha )(1 hb )
+ 2
t 0t 0
tg
k,l=1,...,4
k
(24)
thanks to the form factors

W=

2
L L 2 ut m2 m2 ,
qq W u i dj W
u i dj
2
2
2
16xW (1 xW ) s
kl
Nmn
=
2
Cn
C m C n C m
2 (1 x )2 s 2 dj q k0 u i q k0 dj q l0 u i q l0
xW
W

ut m2u i m2d mn + (m 0 m 0 s)(1 mn ) ,
j
Gmn =
2
2s2 n
ut
Cd q g Cum
q
g
2
i
j
9s

m2u i m2d mn + m2g s (1 mn ) ,
j

2
Re Lqq W Lu i dj W Lu i q 0 Lq q 0 ut m2u i m2d ,
[N W]k = 2
2
2
k
j
j
k
6xW (1 xW ) s

4s
Re Lu i q g Ldj q g Lqq W Lu i dj W ut m2u i m2d ,
[GW] =
j
18xW (1 xW )s 2
(25)
which combine coupling constants and Dirac traces of the squared and interference diagrams.
In cMFV, superpartners of heavy flavours can only be produced through the purely left-handed
s-channel W -exchange, since the t -channel diagrams are suppressed by the small bottom and
negligible top quark densities in the proton, and one recovers the result in Ref. [26]. In NMFV,
t -channel exchanges can, however, contribute to heavy-flavour final state production from lightflavour initial states and even become dominant, due to the strong gluino coupling.
Fig. 2. Tree-level Feynman diagrams for the production of neutral squarkantisquark pairs in quarkantiquark (top) and
gluongluon collisions (bottom).
Neutral squarkantisquark pair production proceeds either from equally neutral quark
antiquark initial states
q(ha , pa )q (hb , pb ) qi (p1 )qj (p2 ),
(26)
through the five different gauge-boson/gaugino exchanges shown in Fig. 2 (top) or from gluon
gluon initial states
g(ha , pa )g(hb , pb ) qi (p1 )qi (p2 )
(27)
through the purely strong couplings shown in Fig. 2 (bottom). The differential cross section for
quarkantiquark scattering

q q
1 [GZ]
1
d ha ,hb
Y
Z1
G
G11 [YZ]1 [GY]
+
= (1 ha )(1 + hb ) 2 + 2 + 2 + 2 +
+
dt
ssz
tg s
tg sz
s
sz
s
tg

kl
k
k

1
N11
[N Y]1 [N Z]1 [N G]k1
[GG]
+
+
+
+
+
tg s
t 0t 0
t 0 s
t 0 sz
t 0 s
k,l=1,...,4 k l
k=1,...,4
k
k
k

2 [GZ]
2
Y
Z2
G
G22 [YZ]2 [GY]
+ (1 + ha )(1 hb ) 2 + 2 + 2 + 2 +
+
+
ssz
tg s
tg sz
s
sz
s
tg

kl

2
N22
[N Y]k2 [N Z]k2 [N G]k2
[GG]
+
+
+
+
+
tg s
t 0t 0
t 0 s
t 0 sz
t 0 s
k,l=1,...,4 k l
k=1,...,4
k
k
k

kl

N12
G12
+ (1 ha )(1 hb ) 2 +
t 0t 0
tg
k,l=1,...,4 k l

kl

N21
G21
+ (1 + ha )(1 + hb ) 2 +
(28)
t 0 t 0
tg
k,l=1,...,4
involves many different form factors,

Y=
2 eq2 eq2 ij qq
s2

ut m2qi m2q ,
j
10
m 2
2
ut
|Lqi qj Z + Rqi qj Z |2 Cqq
Z
2
16s 2 xW (1 xW )2

2s2 ij qq
ut m2qi m2q ,
j
9s 2
Zm =
G=
kl
=
Nmn
2
C m C m C n
C n
2 (1 x )2 s 2 qi q k0 qi q l0 qj q k0 qj q l0
xW
W

ut m2qi m2qj mn + (m 0 m 0 s)(1 mn ) ,
k
Gmn =

m2qi m2q ,
2s2 m n 2
C
ut
C
9s 2 qi q g qj q g

m2qi m2qj mn + m2g s (1 mn ) ,
[YZ]m =

2 eq eq ij qq
m
2 2
Re[Lqi qj Z + Rqi qj Z ]Cqq
Z ut mq mq ,
i
j
2s 2 xW (1 xW )
[N Y]km =

2 2 eq eq ij qq m
2 2
ut
m
,
Re Cq q 0 Cqm
m
0

q
2
i
j
i
3xW (1 xW )s
k j q k
[N Z]km =

m
2
m
2 2
Re Cqm
0 Cq q 0 (Lqi qj Z + Rqi qj Z ) Cqq Z ut mqi mqj ,
2
2
2
i
j
k
k
6xW (1 xW ) s
[N G]km =

8s ij qq
ut m2qi m2qj ,
Re Cqm
C m
i q k0 qj q k0
9xW (1 xW )s 2

4s2 ij qq m m
Re Cqi q g Cqj q g ut m2qi m2qj ,

2
27s

8s eq eq ij qq m m
m=
Re Cqi q g Cqj q g ut m2qi m2qj ,
[GY]
2
9s

m
2
s
m
2 2
m=
Re Cqm
[GZ]
i q g Cqj q g (Lqi qj Z + Rqi qj Z ) Cqq Z ut mqi mqj ,
2
9xW (1 xW )s
m=
[GG]
(29)
since only very few interferences (those between strong and electroweak channels of the same
propagator type) are eliminated due to colour conservation. On the other hand, the gluon-initiated
cross section
gg
d ha ,hb
dt

tqi uqi
s2
8
=
24 1 2 2
3
128s 2
s

sm2qi
sm2qi
(1 ha hb )
(1 ha hb ) 2
tqi uqi
tqi uqi
(30)
involves only the strong coupling constant and is thus quite compact. In the case of cMFV, but
diagonal or non-diagonal squark helicity, our results agree with those in Ref. [26]. Diagonal production of identical squarkantisquark mass eigenstates is, of course, dominated by the strong
quarkantiquark and gluongluon channels. Their relative importance depends on the partonic
luminosity and thus on the type and energy of the hadron collider under consideration. Nondiagonal production of squarks of different helicity or flavour involves only electroweak and
gluino-mediated quarkantiquark scattering, and the relative importance of these processes depends largely on the gluino mass.
11
Fig. 3. Tree-level Feynman diagrams for the production of one down-type squark (qi ) and one up-type squark (qj ) in the
collision of an up-type quark (q) and a down-type quark (q ).
3.3. Squark pair production

While squarkantisquark pairs are readily produced in p p collisions, e.g. at the Tevatron, from
valence quarks and antiquarks, pp colliders have a larger quarkquark luminosity and will thus
more easily lead to squark pair production. The production of one down-type and one up-type
squark
q(ha , pa )q (hb , pb ) di (p1 )u j (p2 ),
(31)
in the collision of an up-type quark q and a down-type quark q proceeds through the t -channel
chargino or u-channel neutralino and gluino exchanges shown in Fig. 3. The corresponding cross
section

qq
kl
C kl

d ha ,hb
N11
G11
11
= (1 ha )(1 hb )
+
+ 2
dt
t t
u 0u 0
ug
k=1,2 k l
k=1,...,4
k

+
l=1,2

k=1,2
l=1,...,4
[CN ]kl
11
tk u 0

k=1,2
l=1,...,4

k=1,2
l=1,...,4

k=1,2
l=1,2

k=1,2
l=1,2

+ (1 + ha )(1 hb )
k=1,...,4
l=1,...,4
kl
N22
u 0 u 0
k

+
G22
u2g
G12
u2g
G21
u2g

k=1,2

[CN ]kl
12
tk u 0
kl
C22
tk tl
[CG]k
22
tk ug
+ (1 ha )(1 + hb )

k=1,2

[CN ]kl
22
tk u 0
[CG]k
11
tk ug
+ (1 + ha )(1 + hb )
l=1,...,4
kl
C12
tk tl
k=1,...,4
l=1,...,4
[CG]k
12
tk ug
kl
N12
u 0 u 0
k

k=1,2

k=1,2
l=1,2
kl
C21
tk tl

+

k=1,...,4
l=1,...,4
kl
N21
u 0 u 0
k
12

+
[CN ]kl
21
tk u 0
k=1,2
l=1,...,4

+
[CG]k
21
tk ug

(32)
k=1,2
involves the form factors

kl
Cmn
=
kl
Nmn
=

2 n
m
n
m
2 2
m smn ,
ut
m
(1
C
C
C
C
m
)
+
m
mn

2 s 2 u j q k di q k u j q l di q l
j
di
k
l
4xW
2
C m C n C m C n
2 (1 x )2 s 2 u j q k0 di q k0 u j q l0 di q l0
xW
W

ut m2d m2u j (1 mn ) + m 0 m 0 smn ,

k
Gmn =
2

2s2 m
Cu j q g Cdn q g ut m2d m2u j (1 mn ) + m2g smn ,
2
i
i
9s
[CN ]kl
mn =

2
m
n
Re Cun q Cdm
i q Cu j q 0 Cdi q 0
2
2
j
k
l
l
k
3xW (1 xW )s

ut m2d m2u j (mn 1) + m m 0 smn ,

k
[CG]kmn =
4s
m
n
Re Cun q Cdm
i q Cu j q g Cdi q g
j
9s 2 xW
k
k

ut m2d m2u j (mn 1) + m mg smn ,

i
(33)
where the neutralinogluino interference term is absent due to colour conservation. The cross
section for the charge-conjugate production of antisquarks from antiquarks can be obtained from
the equations above by replacing ha,b ha,b . Heavy-flavour final states are completely absent
in cMFV due to the negligible top quark and small bottom quark densities in the proton and can
thus only be obtained in NMFV.
The Feynman diagrams for pair production of two up- or down-type squarks
q(ha , pa )q (hb , pb ) qi (p1 )qj (p2 )
(34)
are shown in Fig. 4. In NMFV, neutralino and gluino exchanges can lead to identical squark
flavours for different quark initial states, so that both t - and u-channels contribute and may interfere. The cross section
Fig. 4. Tree-level Feynman diagrams for the production of two up-type or down-type squarks.
13
qq
d ha ,hb
dt

= (1 ha )(1 hb )

k=1,...,4
l=1,...,4
[N U]kl
[N T U]kl
[N T ]kl
11
11
11
+
+
t 0 t 0
u 0 u 0
t 0 u 0
k

+
[GT ]11
tg2

[N GA]k
[N GB]k11
[GU ]11 [GT U]11
1
11
+
+
+
+
2
ug tg
t 0 ug
u 0 tg
1 + ij
ug
k=1,...,4
k
k

[N T ]kl
[N U ]kl
[N T U]kl
[GT ]22
22
22
22
+ (1 + ha )(1 + hb )
+
+
+
t 0 t 0
u 0 u 0
t 0 u 0
tg2
k=1,...,4
l=1,...,4
k=1,...,4
l=1,...,4

[N GA]k
[N GB]k22
1
[GU ]22 [GT U]22
22
+
+
+
+
2
ug tg
t 0 ug
u 0 tg
1 + ij
ug
k=1,...,4
k
k

[N T ]kl
[N U ]kl
[N T U]kl
[GT ]12
12
12
12
+ (1 ha )(1 + hb )
+
+
+
t 0 t 0
u 0 u 0
t 0 u 0
tg2

[N GA]k
[N GB]k12
1
12
+
+
t 0 ug
u 0 tg
1 + ij
k=1,...,4
k
k

[N T ]kl
[N U ]kl
[N T U]kl
[GT ]21
21
21
21
+ (1 + ha )(1 hb )
+
+
+
t 0t 0
u 0 u 0
t 0 u 0
tg2
k=1,...,4
k
l=1,...,4
[GU ]21 [GT U]21

+
+
ug tg
u2g
[N GA]k
[N GB]k21
21
+
t 0 ug
u 0 tg
k=1,...,4

1
1 + ij
depends therefore on the form factors

[N T ]kl
mn =
2
C n C m C n
Cm
2 (1 x )2 s 2 qj q k0 qi q k0 qj q l0 qi q l0
xW
W

ut m2qi m2qj (1 mn ) + m 0 m 0 smn ,

k
2
[N U]kl
C n 0 C m 0 C n 0 C m 0
mn = 2
xW (1 xW )2 s 2 qi q k qj q k qi q l qj q l

ut m2qi m2qj (1 mn ) + m 0 m 0 smn ,

k

2 2
n
n
m
[N T U ]kl
Re Cqm
0 Cq q 0 Cq q 0 Cq q 0
mn =
2
2
2
i
j
i
j
k
k
l
l
3xW (1 xW ) s

2 2
ut mqi mqj (1 mn ) + m 0 m 0 smn ,
k
2

2s2 n
C C m ut m2qi m2qj (1 mn ) + m2g smn ,
[GT ]mn =
9s 2 qj q g qi q g
2

2s2 m
n
ut m2 m2 (1 mn ) + m2 smn ,
C
[GU ]mn =
g Cq q g
q
q
i
j
i
j
9s 2
(35)
14
Fig. 5. Tree-level Feynman diagrams for the associated production of squarks and gauginos.

4s2 m n
n
ut m2qi m2qj (1 mn ) + m2g smn ,
Re Cqi q g Cqj q g Cqm
i q g Cqj q g
2
27s

8s
Re Cqn
C m C m C n
[N GA]kmn = 2
j q k0 qi q k0 qi q g qj q g
9s xW (1 xW )

ut m2qi m2qj (1 mn ) + m 0 mg smn ,
[GT U]mn =
[N GB]kmn

8s
m
n
m
Re Cqn
= 2
0 Cq q 0 Cqj q g Cqi q g
q
i
j
9s xW (1 xW )
k
k

2 2
ut mqi mqj (1 mn ) + m 0 mg smn .
(36)
Gluinos will dominate over neutralino exchanges due to their strong coupling, and the two will
only interfere in the mixed t - and u-channels due to colour conservation. At the LHC, uptype squark pair production should dominate over mixed up-/down-type squark production and
down-type squark pair production, since the proton contains two valence up-quarks and only one
valence down-quark. As before, the charge-conjugate production of antisquark pairs is obtained
by making the replacement ha,b ha,b . If we neglect electroweak contributions as well as
squark flavour and helicity mixing and sum over left- and right-handed squark states, our results
agree with those of Ref. [27].
3.4. Associated production of squarks and gauginos
The associated production of squarks and neutralinos or charginos
q(ha , pa )g(hb , pb ) j (p1 )qi (p2 )
(37)
is a semi-weak process that originates from quarkgluon initial states and has both an s-channel
quark and a t -channel squark contribution. They involve both a quarksquarkgaugino vertex
that can in general be flavour violating. The corresponding Feynman diagrams can be seen in
Fig. 5. The squark-gaugino cross section

qg
d ha ,hb

s uj
=
(1 ha )(1 hb )|Lqi q j |2 + (1 + ha )(1 + hb )|Rqi q j |2
2
dt
s
n s
+
+
+
tj (t + m2qi )
tq2i
(1 ha )|Lqi q j |2 + (1 + ha )|Rqi q j |2
2(ut m2qi m2j )

stqi
tj (tj uqi )
(1 ha )(1 hb )|Lqi q j |2 + (1 + ha )(1 + hb )|Rqi q j |2
stqi
(1 ha )|Lqi q j | + (1 + ha )|Rqi q j |
2
(38)
15
Fig. 6. Tree-level Feynman diagrams for the production of gaugino pairs.
where n = 6xW (1 xW ) for neutralinos and n = 12xW for charginos, is sufficiently compact
to be written without the definition of form factors. Note that the t -channel diagram involves the
coupling of the gluon to scalars and does thus not depend on its helicity hb . The cross section
of the charge-conjugate process can be obtained by taking ha ha . Third-generation squarks
can only be produced in NMFV, preferably through a light (valence) quark in the s-channel. For
non-mixing squarks and gauginos, we agree again with the results of Ref. [27].
3.5. Gaugino pair production
Finally, we consider the purely electroweak production of gaugino pairs
q(ha , pa )q (hb , pb ) i (p1 ) j (p2 )
(39)
from quarkantiquark initial states, where flavour violation can occur via the quarksquark
gaugino vertices in the t - and u-channels (see Fig. 6). However, if it were not for different parton
density weights, summation over complete squark multiplet exchanges would make these channels insensitive to the exchanged squark flavour. Furthermore there are no final state squarks that
could be experimentally tagged. The cross section can be expressed generically as
q q
d ha ,hb
dt
2
2

2
(1 ha )(1 + hb ) QuLL ui uj + QtLL ti tj
2
3s

t
+ 2 Re Qu
LL QLL m i m j s
2
2

2
(1 + ha )(1 hb ) QuRR ui uj + QtRR ti tj
2
3s

t
+ 2 Re Qu
RR QRR m i m j s

2
2

2
(1 + ha )(1 + hb ) QuRL ui uj + QtRL ti tj
2
3s

t
2
2
+ Re Qu
RL QRL ut m i m j
+

2
2

2
(1 ha )(1 hb ) QuLR ui uj + QtLR ti tj
2
3s

t
2
2
+ Re Qu
LR QLR ut m i m j ,
+
i.e. in terms of generalized charges. For i j+ -production, these charges are given by

Qu+
LL
6 Ld q L

Lqq Z OijR
k
dk q j
eq ij qq
i
+
,
s
2xW (1 xW )sz
2xW udk
k=1
(40)
16

Qt+
LL
L
6 L

Lqq Z OijL
u k q j u k q i
eq ij qq
,
s
2xW (1 xW )sz
2xW tu k
k=1
Qu+
RR
Qt+
RR
6 Rd q R

Rqq Z OijL
k
dk q j
eq ij qq
i
=
+
,
s
2xW (1 xW )sz
2xW udk
k=1

R
6 R

Rqq Z OijR
u k q j u k q i
eq ij qq
=
,
s
2xW (1 xW )sz
2xW tu k
k=1
Qu+
LR =
6

Rdk q L
k=1
2xW udk
dk q j
Qt+
LR
L
6 R

u k q j u k q i
2xW tu k
k=1
Qu+
RL
6 Ld q R

k
dk q j
i
2xW udk
k=1
Qt+
RL =
R
6 L

u k q j u k q i
k=1
2xW tu k
(41)
Note that there is no interference between t - and u-channel diagrams due to (electromagnetic)
charge conservation. The cross section for chargino-pair production in e+ e -collisions can be deduced by setting eq el = 1, Lqq Z LeeZ = (2Tl3 2el xW ) and Rqq Z ReeZ = 2el xW .
Neglecting all Yukawa couplings, we can then reproduce the calculations of Ref. [28].
The charges of the chargino-neutralino associated production are given by
Qu+0
LL
Qt+0
LL
6 L

OjL
u k q i u k q j0
i Lqq W
+
,
uu k
2sw
k=1
R
6 L
L 0
Oj i Lqq W
1
dk q i dk q j
=
tdk
2(1 xW )xW
2sw
k=1
1
=
2(1 xW )xW
Qu+0
RR
R
6 R

u k q i u k q j0
1
=
,
uu k
2(1 xW )xW k=1
Qt+0
RR
6 R
R 0

1
dk q i dk q j
=
,
tdk
2(1 xW )xW k=1
Qu+0
LR
L
6 R

u k q i u k q j0
1
=
,
uu k
2(1 xW )xW k=1
Qt+0
LR
6 L
R 0

1
dk q i dk q j
=
,
tdk
2(1 xW )xW k=1
Qu+0
RL
Qt+0
RL
17
R
6 L

u k q i u k q j0
1
=
,
uu k
2(1 xW )xW k=1
6 R
L 0

1
dk q i dk q j
=
.
tdk
2(1 xW )xW k=1
(42)
The charge-conjugate process is again obtained by making the replacement ha,b ha,b in
Eq. (40). In the case of non-mixing squarks with neglected Yukawa couplings, we agree with the
results of Ref. [22], provided we correct a sign in their Eq. (2) as described in Ref. [29].
Finally, the charges for the neutralino pair production are given by
Qu00
LL
Qt00
LL =
Qu00
RR
Qt00
RR =

xW (1 xW ) 1 + ij
1
2sz

xW (1 xW ) 1 + ij
1

xW (1 xW ) 1 + ij
Lqq Z Oij R
uQ k
6 L
k q 0

k q 0 LQ
Q
j
Rqq Z Oij R
6 RQ
k q 0 RQ q 0

i
k
j
uQ k
k=1
Rqq Z Oij L
2sz
,

tQ k
k=1
2sz

6 LQ
k q 0 LQ
k q 0

i
j
k=1
2sz

xW (1 xW ) 1 + ij
1
Lqq Z Oij L
6 R
k q 0

k q 0 RQ
Q
j
i
tQ k
k=1

,

,
Qu00
LR
Qt00
LR

xW (1 xW ) 1 + ij
1

=
xW (1 xW ) 1 + ij
6 RQ
k q 0 LQ q 0

i
k
j
uQ k
k=1
6 L
k q 0

k q 0 RQ
Q
j
i
tQ k
k=1
Qu00
RL =
Qt00
RL

xW (1 xW ) 1 + ij
1

xW (1 xW ) 1 + ij
6 LQ
k q 0 RQ q 0

i
k
j
uQ k
k=1
6 R
k q 0

k q 0 LQ
Q
j
i
k=1
tQ k
(43)
which agrees with the results of Ref. [30] in the case of non-mixing squarks.
3.6. Squark decays
We turn now from SUSY particle production to decay processes and show in Fig. 7 the possible decays of squarks into gauginos and quarks (top) as well as into electroweak gauge bosons
and lighter squarks (bottom). Both processes can in general induce flavour violation. The decay
widths of the former are given by
18
Fig. 7. Tree-level Feynman diagrams for squark decays into gauginos and quarks (top) and into electroweak gauge bosons
and lighter squarks (bottom).
qi 0 qk =
j
2m3qi xW (1 xW )

m2qi m2 0 m2qk |Lqi qk 0 |2 + |Rqi qk 0 |2

4m 0 mqk Re Lqi qk 0 Rq q
j
qi q =
j
4m3qi xW
0
i k j

1/2 m2qi , m2 0 , m2qk ,
k j
4m mqk Re Lqi q Rq q
qi gq
k =
2s
k j
3m3qi xW
(44)

m2qi m2 m2q |Lqi q |2 + |Rqi q |2

i k j
k j
1/2

m2qi , m2 , m2q ,
j
(45)

m2qi m2g m2qk |Lqi qk g |2 + |Rqi qk g |2

4mg mqk Re Lqi qk g Rqi qk g 1/2 m2qi , m2g , m2qk ,
while those of the latter are given by
|Lqi qk Z + Rqi qk Z |2 3/2 m2qi , m2Z , m2qk ,

qi Z qk =
3
2
16mqi mZ xW (1 xW )
|Lqi qk W |2 3/2 m2qi , m2W , m2q .

qi W qk =
3
2
k
16mqi mW xW (1 xW )
(46)
(47)
(48)
The usual Kllen function is

(x, y, z) = x 2 + y 2 + z2 2(xy + yz + zx).
(49)
In cMFV, our results agree with those of Ref. [31].

3.7. Gluino decays
Heavy gluinos can decay strongly into squarks and quarks as shown in Fig. 8. The corresponding decay width

2
g
mg m2qj + m2qk |Lqj qk g |2 + |Rqj qk g |2
qj qk =
3
8mg

+ 4mg mqk Re Lqj qk g Rqj qk g 1/2 m2g , m2qj , m2qk
(50)
can in general also induce flavour violation. In cMFV, our result agrees again with the one of
Ref. [31].
19
Fig. 8. Tree-level Feynman diagram for gluino decays into squarks and quarks.
Fig. 9. Tree-level Feynman diagrams for gaugino decays into squarks and quarks (left) and into lighter gauginos and
electroweak gauge bosons (centre and right).
3.8. Gaugino decays

Heavier gauginos can decay into squarks and quarks as shown in Fig. 9 (left) or into lighter
gauginos and electroweak gauge bosons (Fig. 9 centre and right). The analytical decay widths
are
qj q =
k
3
8m3 xW
i

m2 m2qj + m2q |Lqj q |2 + |Rqj q |2
k i

+ 4m mqk Re Lqj q Rq

j qk i
k i
k i

1/2 m2 , m2qj , m2q
(51)
and
0 W
i
8m3 m2W xW
i
m4 + m4 0 2m4W + m2 m2W + m2 0 m2W 2m2 m2 0

i
2 2
OijL + OijR 12m m2W m 0 Re OijL OijR 1/2 m2 , m2 0 , m2W ,
i
Z
i
(52)
m4 + m4 2m4Z + m2 m2Z + m2 m2Z 2m2 m2
i
j
i
j
i
8m3 m2Z xW (1 xW )
i
2
2

OijL + OijR 12m m2Z m Re OijL OijR 1/2 m2 , m2 , m2Z
i
(53)
for charginos and

0 qj qk =
i

m2 0 m2qj + m2qk |Lqj qk 0 |2 + |Rqj qk 0 |2
i
4m3 0 xW (1 xW )
i

+ 4m 0 mqk Re Lqj qk 0 Rq
i
0
j qk i

1/2 m2 0 , m2qj , m2qk
i
(54)
20
and
0 W
i
8m3 0 m2W xW
i
m4 0 + m4 2m4W + m2 0 m2W + m2 m2W 2m2 0 m2

j
2 2
OijL + OijR 12m 0 m2W m Re OijL OijR 1/2 m2 0 , m2 , m2W ,
j
0 0 Z
i
8m3 0 m2Z xW (1 xW )
i
(55)
m4 0 + m4 0 2m4Z + m2 0 m2Z + m2 0 m2Z 2m2 0 m2 0

i
2
2

Oij L + Oij R 12m 0 m2Z m 0 Re Oij L Oij R 1/2 m2 0 , m2 0 , m2Z
i
(56)
for neutralinos, respectively. Chargino decays into a slepton and a neutrino (lepton and sneutrino) can be deduced from the previous equations by taking the proper limits, i.e. by removing
colour factors and up-type masses in the coupling definitions. Our results agree then with those
of Ref. [32] in the limit of non-mixing sneutrinos. Note that the same simplifications also permit
a verification of our results for squark decays into a gaugino and a quark in Eqs. (44) and (45)
when compared to their leptonic counterparts in Ref. [32].
4. Experimental constraints, scans and benchmarks in NMFV SUSY
In the absence of experimental evidence for supersymmetry, a large variety of data can be
used to constrain the MSSM parameter space. For example, sparticle mass limits can be obtained from searches of charginos (m 85 GeV for heavier sneutrinos at LEP2), neutralinos
1
(m 0 59 GeV in minimal supergravity (mSUGRA) from the combination of LEP2 results),
1
gluinos (mg 195 GeV from CDF), stops (mt1 9596 GeV for neutral- or charged-current
decays from the combination of LEP2 results), and other squarks (mq 300 GeV for gluinos
of equal mass from CDF) at colliders [33]. Note that all of these limits have been obtained assuming constrained minimal flavour violation. For non-minimal flavour violation, rather strong
constraints can be obtained from low-energy, electroweak precision, and cosmological observables. These are discussed in the next subsection, followed by several scans for experimentally
allowed/favoured regions of the constrained MSSM parameter space and the definition of four
NMFV benchmark points/slopes. Finally, we exhibit the corresponding chirality and flavour decomposition of the various squark mass eigenstates.
4.1. Low-energy, electroweak precision, and cosmological constraints
In a rather complete analysis of FCNC constraints more than ten years ago [18], upper limits
from the neutral kaon sector (on mK , , /), on B- (mB ) and D-meson oscillations (mD ),
various rare decays (BR(b s ), BR( e ), BR( e ), and BR( )), and electric
dipole moments (dn and de ) were used to impose constraints on non-minimal flavour mixing
in the squark and slepton sectors. The limit obtained for the absolute value in the left-handed,
down-type squark sector was rather weak (|sb
LL | < 4.426 for varying gluino-to-squark mass
ratio), while the limits for the mixed/right-handed, imaginary or sleptonic parts were already
several orders of magnitude smaller. In the meantime, many of the experimental bounds have
21
Table 1
d d
The 95% probability bounds on |ijk l | obtained in Ref. [34]
kl
12
13
23
ij
LL
LR
RL
RR
1.4 102
9.0 102
1.6 101
9.0 105
1.7 102
4.5 103
9.0 105
1.7 102
6.0 103
9.0 103
7.0 102
2.2 101
been improved or absolute values for the observables have been determined, so that an updated
analysis could be performed [34]. The results for the down-type squark sector are cited in Table 1. As can be seen and as has already been hinted at in the introduction, only mixing between
second- and third-generation squarks can be substantial, and this only in the leftleft or right
right chiral sectors, the latter being disfavoured by its scaling with the soft SUSY-breaking mass.
Independent analyses focusing on this particular sector, i.e. on BR(b s ), BR(b s), and
mBs , have been performed recently by two other groups [35,36] with very similar results.
In our own analysis, we take implicitly into account all of the previously mentioned constraints
ct
by restricting ourselves to the case of only one real NMFV parameter, sb
LL = LL . Allowed
regions for this parameter are then obtained by imposing explicitly a number of low-energy, electroweak precision, and cosmological constraints. We start by imposing the theoretically robust
inclusive branching ratio
BR(b s ) = (3.55 0.26) 104 ,
(57)
obtained from the combined measurements of BaBar, Belle, and CLEO [37], at the 2 -level on
the two-loop QCD/one-loop SUSY calculation [36,38], which affects directly the allowed squark
mixing between the second and third generation.
A second important consequence of NMFV in the MSSM is the generation of large splittings
between squark-mass eigenvalues. The splitting within isospin doublets influences the Z- and
W -boson self-energies at zero-momentum Z,W (0) in the electroweak -parameter
2
= Z (0)/MZ2 W (0)/MW
(58)
2
and consequently the W -boson mass MW and the squared sine of the weak mixing angle sin W .
The latest combined fits of the Z-boson mass, width, pole asymmetry, W -boson and top-quark
mass constrain new physics contributions to T = 0.13 0.11 [33] or
= T = 0.00102 0.00086,
(59)
where we have used (MZ ) = 1/127.918. This value is then imposed at the 2 -level on the
one-loop NMFV and two-loop cMFV SUSY calculation [39].
A third observable sensitive to SUSY loop-contributions is the anomalous magnetic moment
a = (g 2)/2 of the muon, for which recent BNL data and the SM prediction disagree by [33]
a = (22 10) 1010 .
(60)
In our calculation, we take into account the SM and MSSM contributions up to two loops [40,41]
and require them to agree with the region above within two standard deviations.
For cosmological reasons, i.e. in order to have a suitable candidate for non-baryonic cold
dark matter [42], we require the lightest SUSY particle (LSP) to be stable, electrically neutral,
and a colour singlet. The dark matter relic density is then calculated using a modified version
22
of DarkSUSY 4.1 [43], that takes into account the six-dimensional squark helicity and flavour
mixing, and constrained to the region
0.094 < CDM h2 < 0.136
(61)
at 95% (2 ) confidence level. This limit has recently been obtained from the three-year data of
the WMAP satellite, combined with the SDSS and SNLS survey and Baryon Acoustic Oscillation
data and interpreted within an eleven-parameter inflationary model [44], which is more general
than the usual six-parameter vanilla concordance model of cosmology. Note that this range is
well compatible with the older, independently obtained range of 0.094 < CDM h2 < 0.129 [45].
4.2. Scans of the constrained NMFV MSSM parameter space
The above experimental limits are now imposed on the constrained MSSM (cMSSM), or minimal supergravity (mSUGRA), model with five free parameters m0 , m1/2 , tan , A0 , and sgn()
at the grand unification scale. Since our scans of the cMSSM parameter space in the common
scalar mass m0 and the common fermion mass m1/2 depend very little on the trilinear coupling A0 , we set it to zero in the following. Furthermore, we fix a small (10), intermediate (30),
and large (50) value for the ratio of the Higgs vacuum expectation values tan . The impact of
the sign of the off-diagonal Higgs mass parameter is investigated for tan = 10 only, before
we set it to > 0 for tan = 30 and 50 (see below).
With these boundary conditions at the grand unification scale, we solve the renormalization
group equations numerically to two-loop order using the computer programme SPheno 2.2.3 [46]
and compute the soft SUSY-breaking masses at the electroweak scale with the complete oneloop formulas, supplemented by two-loop contributions in the case of the neutral Higgs bosons
and the -parameter. At this point we generalize the squark mass matrices as described in
Section 2 in order to account for flavour mixing in the left-chiral sector of the second- and thirdgeneration squarks, diagonalize these mass matrices, and compute the low-energy, electroweak
precision, and cosmological observables with the computer programmes FeynHiggs 2.5.1 [47]
and DarkSUSY 4.1 [43].
For the masses and widths of the electroweak gauge bosons and the mass of the top quark,
we use the current values of mZ = 91.1876 GeV, mW = 80.403 GeV, mt = 174.2 GeV,
Z = 2.4952 GeV, and W = 2.141 GeV. The CKM-matrix elements are computed using the
parameterization

c12 c13
s12 c13
s13 ei
i
i
V = s12 c23 c12 s23 s13 e
(62)
c12 c23 s12 s23 s13 e
s23 c13 ,
s12 s23 c12 c23 s13 ei c12 s23 s12 c23 s13 ei c23 c13
where sij = sin ij and cij = cos ij relate to the mixing of two specific generations i and j and
is the SM CP -violating complex phase. The numerical values are given by
s12 = 0.2243,
s23 = 0.0413,
s13 = 0.0037,
and
= 1.05.
(63)
The squared sine of the electroweak mixing angle sin W = 1

and the electro
2
2
magnetic fine structure constant = 2GF mW sin W / are calculated in the improved Born
approximation using the world average value of GF = 1.16637 105 GeV2 for Fermis coupling constant [33].
Typical scans of the cMSSM parameter space in m0 and m1/2 with a relatively small value
of tan = 10 and A0 = 0 are shown in Figs. 10 and 11 for < 0 and > 0, respectively.
2
m2W /m2Z
Fig. 10. The (m0 , m1/2 )-planes for tan = 10, A0 = 0 GeV, < 0, and = 0, 0.03, 0.05 and 0.1. We show WMAP (black) favoured as well as b s (blue) and charged
LSP (beige) excluded regions of mSUGRA parameter space in constrained minimal ( = 0) and (non-)minimal ( > 0) flavour violation. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 11. The (m0 , m1/2 )-planes for tan = 10, A0 = 0 GeV, > 0, and = 0, 0.03, 0.05 and 0.1. We show a (grey) and WMAP (black) favoured as well as b s (blue)
and charged LSP (beige) excluded regions of mSUGRA parameter space in constrained minimal ( = 0) and (non-)minimal ( > 0) flavour violation. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
23
24
All experimental limits described in Section 4.1 are imposed at the 2 -level. The b s excluded region depends strongly on flavour mixing, while the regions favoured by g 2 and
the dark matter relic density are quite insensitive to variations of the -parameter. constrains
the parameter space only for heavy scalar masses m0 > 2000 GeV and heavy gaugino masses
m1/2 > 1500 GeV, so that the corresponding excluded regions are not shown here.
The dominant SUSY effects in the calculation of the anomalous magnetic moment of the
muon come from induced quantum loops of a gaugino and a slepton. Squarks contribute only
at the two-loop level. This reduces the dependence on flavour violation in the squark sector
considerably. Furthermore, the region < 0 is disfavoured in all SUSY models, since the oneloop SUSY contributions are approximatively given by [48]

100 GeV 2
SUSY,1-loop
13 1010
tan sgn(),
a
(64)
MSUSY
if all SUSY particles (the relevant ones are the smuon, sneutrino, chargino, and neutralino) have a
common mass MSUSY . Negative values of would then increase, not decrease, the disagreement
between the experimental measurements and the theoretical SM value of a . Furthermore, the
measured b s branching ratio excludes virtually all of the region favoured by the dark matter
relic density, except for very high scalar SUSY masses. We therefore do not consider negative
values of in the rest of this work. As stated above, we have also checked that the shape of the
different regions depends extremely weakly on the trilinear coupling A0 .
In Figs. 12 and 13, we show the (m0 , m1/2 )-planes for larger tan , namely tan = 30 and
tan = 50, and for > 0. The regions which are favoured both by the anomalous magnetic
moment of the muon and by the cold dark matter relic density, and which are not excluded by the
b s measurements, are stringently constrained and do not allow for large flavour violation.
4.3. (c)MFV and NMFV benchmark points and slopes
Restricting ourselves to non-negative values of , we now inspect the (m0 , m1/2 )-planes in
Figs. 1113 for cMSSM scenarios that
are allowed/favoured by low-energy, electroweak precision, and cosmological constraints,
permit non-minimal flavour violation among left-chiral squarks of the second and third generation up to 0.1,
and are at the same time collider-friendly, i.e. have relatively low values of m0 and m1/2 .
Our choices are presented in Table 2, together with the nearest pre-WMAP Snowmass Points
(and Slopes, SPS) [49,50] and the nearest post-WMAP scenarios proposed in Ref. [51]. We also
indicate the relevant cosmological region for each point and attach a model line (slope) to it,
given by
A:
180 GeV m1/2 250 GeV,
m0 = 1936 GeV + 12.9m1/2 ,
B:
400 GeV m1/2 900 GeV,
m0 = 4.93 GeV + 0.229m1/2 ,
C:
500 GeV m1/2 700 GeV,
m0 = 54 GeV + 0.297m1/2 ,
D:
575 GeV m1/2 725 GeV,
m0 = 600 GeV.
(65)
These slopes trace the allowed/favoured regions from lower to higher masses and can, of course,
Fig. 12. The (m0 , m1/2 ) planes for tan = 30, A0 = 0 GeV, > 0, and = 0, 0.03, 0.05 and 0.1. We show a (grey) and WMAP (black) favoured as well as b s (blue)
Fig. 13. The (m0 , m1/2 ) planes for tan = 50, A0 = 0 GeV, > 0, and = 0, 0.03, 0.05 and 0.1. We show a (grey) and WMAP (black) favoured as well as b s (blue)
25
26
Table 2
Benchmark points allowing for flavour violation among the second and third generations for A0 = 0, > 0, and three different values of tan . For comparison we also show the nearest pre-WMAP SPS [49,50] and post-WMAP BDEGOP [51]
benchmark points and indicate the relevant cosmological regions
A
B
C
D
m0 [GeV]
m1/2 [GeV]
A0 [GeV]
tan
sgn()
SPS
BDEGOP
Cosmol. region
700
100
230
600
200
400
590
700
0
0
0
0
10
10
30
50
1
1
1
1
2
3
1b
4
E
C
I
L
Focus Point
Co-Annihilation
Co-Annihilation
Bulk/Higgs-funnel
also be used in cMFV scenarios with = 0. We have verified that in the case of MFV [7] the
qq
qq
qq
ct
hierarchy LL LR,RL RR and the equality of sb
LL = LL are still reasonably well fulct
3 . . . 1 102 for
filled numerically with the values of sb
LL LL ranging from zero to 5 10
our four typical benchmark points.
Starting with Fig. 11 and tan = 10, the bulk region of equally low scalar and fermion
masses is all but excluded by the b s branching ratio. This leaves as a favoured region
first the so-called focus point region of low fermion masses m1/2 , where the lightest neutralinos are relatively heavy, have a significant higgsino component, and annihilate dominantly into
pairs of electroweak gauge bosons. Our benchmark point A lies in this region, albeit at smaller
masses than SPS 2 (m0 = 1450 GeV, m1/2 = 300 GeV) and BDEGOP E (m0 = 1530 GeV,
m1/2 = 300 GeV), which lie outside the region favoured by a (grey-shaded) and lead to colliderunfriendly squark and gaugino masses.
The second favoured region for small tan is the co-annihilation branch of low scalar masses
m0 , where the lighter tauslepton mass eigenstate is not much heavier than the lightest neutralino
and the two have a considerable co-annihilation cross section. This is where we have chosen our
benchmark point B, which differs from the points SPS 3 (m0 = 90 GeV, m1/2 = 400 GeV) and
BDEGOP C (m0 = 85 GeV, m1/2 = 400 GeV) only very little in the scalar mass. This minor
difference may be traced to the fact that we use DarkSUSY 4.1 [43] instead of the private dark
matter programme SSARD of Ref. [51].
At the larger value of tan = 30 in Fig. 12, only the co-annihilation region survives the
constraints coming from b s decays. Here we choose our point C, which has slightly higher
masses than both SPS 1b (m0 = 200 GeV, m1/2 = 400 GeV) and BDEGOP I (m0 = 175 GeV,
m1/2 = 350 GeV), due to the ever more stringent constraints from the above-mentioned rare
B-decay.
For the very large value of tan = 50 in Fig. 13, the bulk region reappears at relatively
heavy scalar and fermion masses. Here, the couplings of the heavier scalar and pseudo-scalar
higgses H 0 and A0 to bottom quarks and tau-leptons and the charged-Higgs coupling to topbottom pairs are significantly enhanced, resulting e.g. in increased dark matter annihilation
cross sections through s-channel Higgs-exchange into bottom-quark final states. So as tan
increases further, the so-called Higgs-funnel region eventually makes its appearance on the diagonal of large scalar and fermion masses. We choose our point D in the concentrated (bulky)
region favoured by cosmology and a at masses, that are slightly higher than those of SPS 4
(m0 = 400 GeV, m1/2 = 300 GeV) and BDEGOP L (m0 = 300 GeV, m1/2 = 450 GeV). We do
so in order to escape again from the constraints of the b s decay, which are stronger today
than they were a few years ago. In this scenario, squarks and gluinos are very heavy with masses
above 1 TeV.
27
4.4. Dependence of precision observables and squark-mass eigenvalues on flavour violation

Let us now turn to the dependence of the precision variables discussed in Section 4.1 on the
flavour violating parameter in the four benchmark scenarios defined in Section 4.3. As already
mentioned, we expect the leptonic observable a to depend weakly (at two loops only) on the
squark sector, and this is confirmed by our numerical analysis. We find constant values of 6, 14,
16, and 13 1010 for the benchmarks A, B, C, and D, all of which lie well within 2 (the latter
three even within 1 ) of the experimentally favoured range (22 10) 1010 .
The electroweak precision observable is shown first in Figs. 1417 for the four benchmark
scenarios A, B, C, and D. On our logarithmic scale, only the experimental upper bound of the
2 -range is visible as a dashed line. While the self-energy diagrams of the electroweak gauge
bosons depend obviously strongly on the helicities, flavours, and mass eigenvalues of the squarks
in the loop, the SUSY masses in our scenarios are sufficiently small and the experimental error is
still sufficiently large to allow for relatively large values of 0.57, 0.52, 0.38, and 0.32 for the
benchmark points A, B, C, and D, respectively. As mentioned above, conversely constrains
SUSY models in cMFV only for masses above 2000 GeV for m0 and 1500 GeV for m1/2 .
The next diagram in Figs. 1417 shows the dependence of the most stringent low-energy
constraint, coming from the good agreement between the measured b s branching ratio and
the two-loop SM prediction, on the NMFV parameter . The dashed lines of the 2 -bands exhibit
two allowed regions, one close to = 0 (vertical green line) and a second one around 0.57,
0.75, 0.62, and 0.57, respectively. As is well-known, the latter are, however, disfavoured by
b s+ data constraining the sign of the b s amplitude to be the same as in the SM
[52]. We will therefore limit ourselves later to the regions 0.05 (points A, C, and D) and
0.1 (point B) in the vicinity of (c)MFV (see also Table 1).
The 95% confidence-level (or 2 ) region for the cold dark matter density was given in absolute
values in Ref. [44] and is shown as a dashed band in the upper right part of Figs. 1417. However,
only the lower bound (0.094) is of relevance, as the relic density falls with increasing . This
is not so pronounced in our model B as in our model A, where squark masses are light and
the lightest neutralino has a sizable higgsino-component, so that squark exchanges contribute
significantly to the annihilation cross sections. For models C and D there is little sensitivity of
CDM h2 (except at very large 1), as the squark masses are generally larger.
The rapid fall-off of the relic density for very large 1 can be understood by looking
at the resulting lightest up- and down-type squark mass eigenvalues in the lower left part of
Figs. 1417. For maximal flavour violation, the off-diagonal squark mass matrix elements are
of similar size as the diagonal ones, leading to one squark mass eigenvalue that approaches
and finally falls below the lightest neutralino (dark matter) mass. Light squark propagators and
co-annihilation processes thus lead to a rapidly falling dark matter relic density and finally to
cosmologically excluded NMFV SUSY models, since the LSP must be electrically neutral and a
colour singlet.
An interesting phenomenon of level reordering between neighbouring states can be observed
in the lower central diagrams of Figs. 1417 for the two lowest mass eigenvalues of up-type
squarks. The squark mass eigenstates are, by definition, labeled in ascending order with the mass
eigenvalues, so that u 1 represents the lightest, u 2 the second-lightest, and u 6 the heaviest uptype squark. As and the off-diagonal entries in the mass matrix increase, the splitting between
the lightest and highest mass eigenvalues naturally increases, whereas the intermediate squark
masses (of u 3,4,5 ) are practically degenerate and insensitive to . These remarks also hold for the
down-type squark masses shown in the lower right diagrams of Figs. 1417. However, for up-type
28
Fig. 14. Dependence of the precision variables BR(b s ), , and the cold dark matter relic density CDM h2 (top) as well as of the lightest SUSY particle, up- and down-type
squark masses (bottom) on the NMFV parameter in our benchmark scenario A. The experimentally allowed ranges (within 2 ) are indicated by horizontal dashed lines.
Fig. 15. Same as Fig. 14 for our benchmark scenario B.

29
30
Fig. 16. Same as Fig. 14 for our benchmark scenario C.
Fig. 17. Same as Fig. 14 for our benchmark scenario D.

31
32
squarks it is first the second-lowest mass that decreases up to intermediate values of = 0.20.5,
whereas the lowest mass is constant, and only at this point the second-lowest mass becomes constant and takes approximately the value of the until here lowest squark mass, whereas the lowest
squark mass starts to decrease further with . These avoided crossings are a common phenomenon for Hermitian matrices and reminiscent of meta-stable systems in quantum mechanics. At
the point where the two levels should cross, the corresponding squark eigenstates mix and change
character, as will be explained in the next subsection. For scenario C (Fig. 16), the phenomenon
occurs even a second time with an additional avoided crossing between the states u 2 and u 3 at
0.05. For scenario B (Fig. 15), this takes place at 0.1, and there is even another crossing
at 0.02. For down-type squarks, the level-reordering phenomenon is not so pronounced.
4.5. Chirality and flavour decomposition of squark mass eigenstates
In NMFV, squarks will not only exhibit the traditional mixing of left- and right-handed helicities of third-generation flavour eigenstates, but will in addition exhibit generational mixing. As
discussed before, we restrict ourselves here to the simultaneous mixing of left-handed secondand third-generation up- and down-type squarks. For our benchmark scenario A, the helicity and
flavour decomposition of the six up-type (left) and down-type (right) squark mass eigenstates is
shown in Fig. 18 for the full range of the parameter [0; 1] and in Fig. 19 for the experimentally favoured range in the vicinity of (c)MFV, [0; 0.1]. First-generation and right-handed
second-generation squarks remain, of course, helicity- and flavour-diagonal,
u 4 = u R ,
d2 = sR ,
d3 = dR ,
u 5 = u L ,
d5 = dL ,
u 3 = cR ,
(66)
with the left-handed and first-generation squarks being slightly heavier due their weak isospin
coupling (see Eqs. (5)(7)) and different renormalization-group running effects. Production of
these states will benefit from t - and u-channel contributions of first- and second-generation
quarks with enhanced parton densities in the external hadrons, but they will not be identified
easily with heavy-flavour tagging and are of little interest for our study of flavour violation. The
lightest up-type squark u 1 remains the traditional mixture of left- and right-handed stops over
a large region of 0.4, but it shows at this point an interesting flavour transition, which is
in fact expected from the level reordering phenomenon discussed in the lower central plot of
Fig. 14. The transition happens, however, above the experimental limit of 0.1. Below this
limit, it is the states u 2 , u 6 , d1 , and in particular d4 and d6 that show, in addition to helicity mixing, the most interesting and smooth variation of second- and third-generation flavour content
(see Fig. 19). Note that at very low 0.002 the states dL and sL rapidly switch levels. This
numerically small change was not visible on the linear scale in Fig. 14.
For the benchmark point B, whose helicity and flavour decomposition is shown in Fig. 20,
level reordering occurs at 0.4 for the intermediate-mass up-type squarks,
u 4,3 = u R ,
d2 = sR ,
d3 = dR ,
u 5 = u L ,
d5 = dL
u 3,4 = cR ,
(67)
whereas the ordering of down-type squarks is very similar to scenario A. Close inspection of
Fig. 21 shows, however, that also dR and sR switch levels at low values of 0.02. At

33
Fig. 18. Dependence of the chirality (L, R) and flavour (u, c, t ; d, s, and b) content of up- (u i ) and down-type (di ) squark mass eigenstates on the NMFV parameter [0; 1]
for benchmark point A.
Fig. 19. Same as Fig. 18 for [0; 0.1].
34
35
Fig. 20. Same as Fig. 18 for benchmark point B.
36
37
Fig. 22. Same as Fig. 18 for benchmark point C.
38
39
Fig. 24. Same as Fig. 18 for benchmark point D.
40
41
0.01, in addition sR and bL switch levels, and at 0.002 it is the states u L and cL . The
lightest up-type squark is again nothing but a mix of left- and right-handed stops up to
0.4. Phenomenologically smooth transitions below 0.1 involving taggable third-generation
squarks are observed for u 4 , u 6 , d1 , and d6 .
The helicity and flavour decomposition for our scenario D, shown in Fig. 24, is rather similar
to the one in scenario A, i.e.
u 4 = u R ,
d3 = sR ,
d4 = dR ,
u 5 = u L ,
d5 = dL
u 3 = cR ,
(68)
are exactly the same in the up-squark sector, and only the mixed down-type state d4 is now lighter
and becomes d2 . The lightest up-type squark, u 1 , is again mostly a mix of left- and right-handed
top squarks up to 0.4, where the level reordering and generation mixing occurs (see lower
central part of Fig. 17). At the experimentally favoured lower values of 0.1, the states u 2 ,
u 6 , d1 , d2 , and d6 exhibit some smooth variations, shown in detail in Fig. 25, albeit to a lesser
extent than in scenario A. At very low 0.004, it is now the up-type squarks u L and cL that
rapidly switch levels. This numerically small change was again not visible on a linear scale (see
Fig. 17).
For our scenario C, shown in Fig. 22, the assignment of the intermediate states
u 4 = u R ,
d3 = sR ,
d4 = dR ,
u 5 = u L ,
d5 = dL
u 3 = cR ,
(69)
is the same as for scenario D above 0.1. Just below, u R and cR as well as dR and sR rapidly
switch levels, whereas u L and cL switch levels at very low 0.002. These changes were
already visible upon close inspection of the lower central and right plots in Fig. 16. On the other
hand, the lightest squarks u 1 and d1 only acquire significant flavour admixtures at relatively large
0.20.4, whereas they are mostly superpositions of left- and right-handed stops and sbottoms
in the experimentally favourable range of 0.1 shown in Fig. 23. Here, the heaviest states u 6
and d6 show already smooth admixtures of third-generation squarks as it was the case for the
scenarios A and D discussed above. The most interesting states are, however, u 2 , u 4 , d2 , and d4 ,
respectively, since they represent combinations of up to four different helicity and flavour states
and have a significant, taggable third-generation flavour content.
5. Numerical predictions for NMFV SUSY particle production at the LHC
In this section, we present numerical predictions for the production cross sections of squark
antisquark pairs, squark pairs, the associated production of squarks
and gauginos, and gaugino
pairs in NMFV SUSY at the CERN LHC, i.e. for pp-collisions at S = 14 TeV centre-of-mass
energy. Thanks to the QCD factorization theorem, total unpolarized hadronic cross sections
1
=
4m2 /S
1/2
ln
d
1/2 ln
tmax
d
dy
dt fa/A xa , Ma2 fb/B xb , Mb2
dt
tmin
(70)
42
can be calculated by convolving the relevant partonic cross sections d /dt, computed in SecA, B, which
tion 3, with universal parton densities fa/A and fb/B of partons a, b in the hadrons
depend on the longitudinal momentum fractions of the two partons xa,b = ey and on the
unphysical factorization scales Ma,b . For consistency with our leading order (LO) QCD calculation in the collinear approximation, where all squared quark masses (except for the top-quark
mass) m2q s, we employ the LO set of the latest CTEQ6 global parton density fit [53], which
includes nf = 5 light (including the bottom) quark flavours and the gluon, but no top-quark
density. Whenever it occurs, i.e. for gluon initial states and gluon or gluino exchanges, the strong
nf =5
= 165 MeV.
coupling constant s (R ) is calculated with the corresponding LO value of LO
We identify the renormalization scale R with the factorization scales Ma = Mb and set the
scales to the average mass of the final state SUSY particles i and j , m = (mi + mj )/2.
The numerical cross sections for charged squarkantisquark and squarksquark production,
neutral up- and down-type squarkantisquark and squarksquark pair production, associated production of squarks with charginos and neutralinos, and gaugino pair production are shown in
Fig. 26 for our benchmark scenario A, in Fig. 27 for scenario B, in Fig. 28 for scenario C, and
in Fig. 29 for scenario D. The magnitudes of the cross sections vary from the barely visible level
of 102 fb for weak production of heavy final states over the semi-strong production of average
squarks and gauginos and quarkgluon initial states to large cross sections of 102 to 103 fb for the
strong production of diagonal squark(anti)squark pairs or weak production of very light gaugino
pairs. Unfortunately, these processes, whose cross sections are largest (top right, center left, and
lower right parts of Figs. 2629), are practically insensitive to the flavour violation parameter ,
as the strong gauge interaction is insensitive to quark flavours and gaugino pair production cross
sections are summed over exchanged squark flavours.
Some of the subleading, non-diagonal cross sections show, however, sharp transitions, in
particular down-type squarkantisquark production at the benchmark point B (centre-left part
of Fig. 27), but also the other squarkantisquark and squarksquark production processes. At
= 0.02, the cross sections for d1 d6 and d3 d6 switch places. Since the concerned sstrange and
sbottom mass differences are rather small, this is mainly due to the different strange and bottom
quark densities in the proton. The cross section is mainly due to the exchange of strongly coupled gluinos despite their larger mass. At = 0.035 the cross sections for d3 d6 and d1 d3 increase
sharply, since d3 = dR can then be produced from down-type valence quarks. The cross section
of the latter process increases with the strange squark content of d1 .
At the benchmark point C (Fig. 28), sharp transitions occur between the u 2 /u 4 and d2 /d4
states, which are pure charm/strange squarks below/above = 0.035, for all types of charged
and neutral squarkantisquark and squarksquark production and also squark-gaugino associated
production. As a side-remark we note that an interesting perspective might be the exploitation
of these t -channel contributions to second- and third-generation squark production for the determination of heavy-quark densities in the proton. This requires, of course, efficient experimental
techniques for heavy-flavour tagging.
Smooth transitions and semi-strong cross sections of about 1 fb are observed for the associated production of third-generation squarks with charginos (lower left diagrams) and neutralinos
(lower centre diagrams) and in particular for the scenarios A and B. For benchmark point A
(Fig. 26), the cross section for d4 production decreases with its strange squark content, while
the bottom squark content increases at the same time. For benchmark point B (Fig. 27), the
same (opposite) happens for d6 (d1 ), while the cross sections for u 6 increase/decrease with its
charm/top squark content. Even in minimal flavour violation, the associated production of stops
and charginos is a particularly useful channel for SUSY particle spectroscopy, as can be seen

Fig. 26. Cross sections for charged squarkantisquark (top left) and squarksquark (top centre) production, neutral up-type (top right) and down-type (centre left)
squarkantisquark and squarksquark pair (centre and centre right) production, associated production of squarks with charginos (bottom left) and neutralinos (bottom centre), and gaugino pair production (bottom right) at the LHC in our benchmark scenario A.
43
Fig. 26. (continued)
44
Fig. 27. Same as Fig. 26 for our benchmark scenario B.
45
46
Fig. 28. Same as Fig. 26 for our benchmark scenario C.
47
48
Fig. 29. Same as Fig. 26 for our benchmark scenario D.
49
50
51
Table 3
Dominant s-, t -, and u-channel contributions to the flavour violating hadroproduction of third-generation squarks and/or
gauginos and the competing dominant flavour-diagonal contributions
Final state
Exchange
s
W
NMFV-Z
NMFV-Z
NMFV-g
NMFV-g
NMFV-g
g
NMFV-g
b
NMFV-b
NMFV-t
t
q
NMFV-g
g
NMFV-g
tb
s
b
t c
tb
b b
tt
0 b
b
0 t
t

b
NMFV-c
NMFV-c
b
, Z, W
from the fact that cross sections vary over several orders of magnitude among our four benchmark points (see also Ref. [54]).
An illustrative summary of flavour violating hadroproduction cross section contributions for
third-generation squarks and/or gauginos is presented in Table 3, together with the competing
flavour-diagonal contributions.
6. Conclusions
In conclusion, we have performed an extensive analysis of squark and gaugino hadroproduction and decays in non-minimal flavour violating supersymmetry. Within the super-CKM basis,
we have taken into account the possible misalignment of quark and squark rotations and computed all squared helicity amplitudes for the production and the decay widths of squarks and
gauginos in compact analytic form, verifying that our results agree with the literature in the
case of non-mixing squarks whenever possible. Flavour violating effects have also been included
in our analysis of dark matter (co-)annihilation processes. We have then analyzed the NMFV
SUSY parameter space for regions allowed by low-energy, electroweak precision, and cosmological data and defined four new post-WMAP benchmark points and slopes equally valid in
minimal and non-minimal flavour violating SUSY. We found that left-chiral mixing of secondand third-generation squarks is slightly stronger constrained than previously believed, mostly
due to smaller experimental errors on the b s branching ratio and the cold dark matter relic
density. For our four benchmark points, we have presented the dependence of squark mass eigenvalues and the flavour and helicity decomposition of the squark mass eigenstates on the flavour
violating parameter . We have computed numerically all production cross sections for the LHC
and discussed in detail their dependence on flavour violation. A full experimental study including heavy-flavour tagging efficiencies, detector resolutions, and background processes would, of
course, be very interesting in order to establish the experimental significance of NMFV. While
the implementation of our analytical results in a general-purpose Monte Carlo generator should
now be straightforward, such a detailed experimental study represents a research project of its
own [55] and is beyond the scope of the work presented here.
52
Acknowledgements
A large part of this work has been performed in the context of the CERN 2006/2007 workshop
on Flavour in the Era of the LHC. The authors also acknowledge interesting discussions with
J. Debove, A. Djouadi, W. Porod, J.M. Richard, and P. Skands. This work was supported by two
Ph.D. fellowships of the French ministry for education and research.
Appendix A. Gaugino and higgsino mixing
The soft SUSY-breaking terms in the minimally supersymmetric Lagrangian include a term
1
0 T
Y 0 + h.c.,
2
which is bilinear in the (2-component) fermionic partners

i W 3 , H 0 , H 0 T
0 = i B,
L
(A.1)
(A.2)
of the neutral electroweak gauge and Higgs bosons and proportional to the, generally complex
and symmetric, neutralino mass matrix
M1
0
mZ sW c
mZ sW s
0
M2
mZ cW c mZ cW s
Y =
(A.3)
.
mZ sW c mZ cW c
0
mZ sW s
mZ cW s
0
Here, M1 , M2 , and are the SUSY-breaking bino, wino, and off-diagonal higgsino mass parameters with tan = s /c = vu /vd being the ratio of the vacuum expectation values vu,d of the
two Higgs doublets, while mZ is the SM Z-boson mass and sW (cW ) is the sine (co-sine) of the
electroweak mixing angle W . After electroweak gauge-symmetry breaking and diagonalization
of the mass matrix Y , one obtains the neutralino mass eigenstates
i0 = Nij j0 ,
i, j = 1, . . . , 4,
(A.4)
where N is a unitary matrix satisfying the relation

N Y N 1 = diag(m 0 , m 0 , m 0 , m 0 ).
1
(A.5)
In 4-component notation, the Majorana-fermionic neutralino mass eigenstates can be written as

0
i
0
i =
(A.6)
, i = 1, . . . , 4.
i0
Their mass eigenvalues m 0 can, e.g., be found in analytic form in [56] and can be chosen to be
i
real and non-negative.
The chargino mass term in the SUSY Lagrangian

+
0 XT
1
L +
(A.7)
+ h.c.
X 0
2
is bilinear in the (2-component) fermionic partners
T
j = i W , H 2,1
(A.8)
53
of the charged electroweak gauge and Higgs bosons and proportional to the, generally complex,
chargino mass matrix

mW 2s
M
2
,
X=
(A.9)
mW 2c
where mW is the mass of the SM W -boson. Its diagonalization leads to the chargino mass eigenstates
i+ = Vij j+ ,
i = Uij j ,
i, j = 1, 2,
(A.10)
where the matrices U and V satisfy the relation

U XV 1 = diag(m , m ).
1
(A.11)
In 4-component notation, the Dirac-fermionic chargino mass eigenstates can be written as

i
i =
(A.12)
.
i
The mass eigenvalues can be chosen to be real and non-negative and are given by [2]
m2 =
1,2
2

1 2
M + 2 + 2m2W M22 2
2 2
1/2
2
,
+ 4m4W c2
+ 4m2W M22 + 2 + 2M2 s2
while the matrices

U = O
and V =
O+ ,
3 O+ ,
if det X 0
if det X < 0
(A.13)

with O =
cos
sin
are determined by the mixing angles with 0 /2 and
2 2mW (M2 sin + cos )

and
tan 2+ =
M22 2 + 2m2W cos 2
2 2mW (M2 cos + sin )

.
tan 2 =
M22 2 2m2W cos 2
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
H.P. Nilles, Phys. Rep. 110 (1984) 1.

H.E. Haber, G.L. Kane, Phys. Rep. 117 (1985) 75.
M. Ciuchini, G. Degrassi, P. Gambino, G.F. Giudice, Nucl. Phys. B 534 (1998) 3.
A.J. Buras, P. Gambino, M. Gorbahn, S. Jger, L. Silvestrini, Phys. Lett. B 500 (2001) 161.
L.J. Hall, L. Randall, Phys. Rev. Lett. 65 (1990) 2939.
G. DAmbrosio, G.F. Giudice, G. Isidori, A. Strumia, Nucl. Phys. B 645 (2002) 155.
A.J. Buras, P. Gambino, M. Gorbahn, S. Jager, L. Silvestrini, Phys. Lett. B 500 (2001) 161;
W. Altmannshofer, A.J. Buras, D. Guadagnoli, hep-ph/0703200.
N. Cabibbo, Phys. Rev. Lett. 10 (1963) 531.
M. Kobayashi, T. Maskawa, Prog. Theor. Phys. 49 (1973) 652.
L.J. Hall, V.A. Kostelecky, S. Raby, Nucl. Phys. B 267 (1986) 415.
J.F. Donoghue, H.P. Nilles, D. Wyler, Phys. Lett. B 128 (1983) 55.
M.J. Duncan, Nucl. Phys. B 221 (1983) 285.
sin
cos

(A.14)
(A.15)
54
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
A. Bouquet, J. Kaplan, C.A. Savoy, Phys. Lett. B 148 (1984) 69.

F. Borzumati, A. Masiero, Phys. Rev. Lett. 57 (1986) 961.
F. Gabbiani, A. Masiero, Nucl. Phys. B 322 (1989) 235.
P. Brax, C.A. Savoy, Nucl. Phys. B 447 (1995) 227.
J.S. Hagelin, S. Kelley, T. Tanaka, Nucl. Phys. B 415 (1994) 293.
F. Gabbiani, E. Gabrielli, A. Masiero, L. Silvestrini, Nucl. Phys. B 477 (1996) 321.
M. Ciuchini, E. Franco, D. Guadagnoli, V. Lubicz, M. Pierini, V. Porretti, L. Silvestrini, hep-ph/0703204.
W. Beenakker, R. Hpker, M. Spira, P.M. Zerwas, Nucl. Phys. B 492 (1997) 51;
E.L. Berger, M. Klasen, T. Tait, Phys. Rev. D 59 (1999) 074024.
H. Baer, B.W. Harris, M.H. Reno, Phys. Rev. D 57 (1998) 5871.
W. Beenakker, M. Klasen, M. Krmer, T. Plehn, M. Spira, P.M. Zerwas, Phys. Rev. Lett. 83 (1999) 3780.
E.L. Berger, M. Klasen, T. Tait, Phys. Lett. B 459 (1999) 165;
E.L. Berger, M. Klasen, T.M.P. Tait, Phys. Rev. D 62 (2000) 095014;
E.L. Berger, M. Klasen, T.M.P. Tait, Phys. Rev. D 67 (2003) 099901, Erratum.
W. Beenakker, M. Krmer, T. Plehn, M. Spira, P.M. Zerwas, Nucl. Phys. B 515 (1998) 3.
E.L. Berger, B.W. Harris, D.E. Kaplan, Z. Sullivan, T.M.P. Tait, C.E.M. Wagner, Phys. Rev. Lett. 86 (2001) 4231.
G. Bozzi, B. Fuks, M. Klasen, Phys. Rev. D 72 (2005) 035016.
S. Dawson, E. Eichten, C. Quigg, Phys. Rev. D 31 (1985) 1581.
S.Y. Choi, A. Djouadi, H.S. Song, P.M. Zerwas, Eur. Phys. J. C 8 (1999) 669.
M. Spira, Nucl. Phys. B (Proc. Suppl.) 89 (2000) 222.
G.J. Gounaris, J. Layssac, P.I. Porfyriadis, F.M. Renard, Phys. Rev. D 70 (2004) 033011.
A. Bartl, W. Majerotto, W. Porod, Z. Phys. C 64 (1994) 499;
A. Bartl, W. Majerotto, W. Porod, Z. Phys. C 68 (1995) 518.
M. Obara, N. Oshimo, JHEP 0608 (2006) 054.
W.M. Yao, et al., Particle Data Group, J. Phys. G 33 (2006) 1.
M. Ciuchini, A. Masiero, P. Paradisi, L. Silvestrini, S.K. Vempati, O. Vives, hep-ph/0702144.
J. Foster, K.I. Okumura, L. Roszkowski, Phys. Lett. B 641 (2006) 452.
T. Hahn, W. Hollik, J.I. Illana, S. Penaranda, hep-ph/0512315.
E. Barberio, et al., Heavy Flavor Averaging Group, hep-ex/0603003.
A.L. Kagan, M. Neubert, Phys. Rev. D 58 (1998) 094012.
S. Heinemeyer, W. Hollik, F. Merz, S. Penaranda, Eur. Phys. J. C 37 (2004) 481.
S. Heinemeyer, D. Stckinger, G. Weiglein, Nucl. Phys. B 690 (2004) 62.
S. Heinemeyer, D. Stckinger, G. Weiglein, Nucl. Phys. B 699 (2004) 103.
J.R. Ellis, J.S. Hagelin, D.V. Nanopoulos, K.A. Olive, M. Srednicki, Nucl. Phys. B 238 (1984) 453.
P. Gondolo, J. Edsjo, P. Ullio, L. Bergstrom, M. Schelke, E.A. Baltz, JCAP 0407 (2004) 008.
J. Hamann, S. Hannestad, M.S. Sloth, Y.Y.Y. Wong, Phys. Rev. D 75 (2007) 023522.
J.R. Ellis, K.A. Olive, Y. Santoso, V.C. Spanos, Phys. Lett. B 565 (2003) 176.
W. Porod, Comput. Phys. Commun. 153 (2003) 275.
S. Heinemeyer, W. Hollik, G. Weiglein, Comput. Phys. Commun. 124 (2000) 76.
T. Moroi, Phys. Rev. D 53 (1996) 6565;
T. Moroi, Phys. Rev. D 56 (1997) 4424, Erratum.
B.C. Allanach, et al., Eur. Phys. J. C 25 (2002) 113.
J.A. Aguilar-Saavedra, et al., Eur. Phys. J. C 46 (2006) 43.
M. Battaglia, A. De Roeck, J.R. Ellis, F. Gianotti, K.A. Olive, L. Pape, Eur. Phys. J. C 33 (2004) 273.
P. Gambino, U. Haisch, M. Misiak, Phys. Rev. Lett. 94 (2005) 061803.
J. Pumplin, D.R. Stump, J. Huston, H.L. Lai, P. Nadolsky, W.K. Tung, JHEP 0207 (2002) 012.
M. Beccaria, G. Macorini, L. Panizzi, F.M. Renard, C. Verzegnassi, Phys. Rev. D 74 (2006) 093009.
B. Fuks, M. Klasen, P. Skands, in preparation.
M.M. El Kheishen, A.A. Aboshousha, A.A. Shafik, Phys. Rev. D 45 (1992) 4345.
Complex marginal deformations of D3-brane

geometries, their Penrose limits and giant gravitons
Spyros D. Avramis a,b, , Konstadinos Sfetsos a , Dimitrios Zoakos a
a Department of Engineering Sciences, University of Patras, 26110 Patras, Greece
b Department of Physics, National Technical University of Athens, 15773 Athens, Greece
Received 27 April 2007; accepted 6 July 2007

Available online 31 July 2007
Abstract
We apply the LuninMaldacena construction of gravity duals to -deformed gauge theories to a class of
type IIB backgrounds with U (1)3 global symmetry, which include the multicenter D3-brane backgrounds
dual to the Coulomb branch of N = 4 super-YangMills and the rotating D3-brane backgrounds dual to
the theory at finite temperature and chemical potential. After a general discussion, we present the full form
of the deformed metrics for three special cases, which can be used for the study of various aspects of
the marginally-deformed gauge theories. We also construct the Penrose limits of the solutions dual to the
Coulomb branch along a certain set of geodesics and, for the resulting PP-wave metrics, we examine the
effect of -deformations on the giant graviton states. We find that giant gravitons exist only up to a critical
value of the -deformation parameter, are not degenerate in energy with the point graviton, and remain
perturbatively stable. Finally, we probe the -deformed multicenter solutions by examining the static heavyquark potential by means of Wilson loops. We find situations that give rise to complete screening as well as
linear confinement, with the latter arising is an intriguing way reminiscent of phase transitions in statistical
systems.
1. Introduction
The AdS/CFT correspondence [1] has proven to be an invaluable tool for exploring the dynamics of large N gauge theories at strong coupling. In its original form, it relates N = 4, SU(N )
* Corresponding author at: Department of Engineering Sciences, University of Patras, 26110 Patras, Greece.
E-mail addresses: avramis@mail.cern.ch (S.D. Avramis), sfetsos@upatras.gr (K. Sfetsos), dzoakos@upatras.gr

(D. Zoakos).
doi:10.1016/j.nuclphysb.2007.07.017
56
S.D. Avramis et al. / Nuclear Physics B 787 (2007) 5597
super-YangMills theory to type IIB string theory on AdS5 S5 , with the limit of large t Hooft
coupling in the gauge theory corresponding to the classical supergravity limit of the string theory.
The AdS/CFT correspondence can be extended to less symmetric theories, a class of which is
the exactly marginal deformations of N = 4 SYM, introduced by Leigh and Strassler [2], which
break supersymmetry down to N = 1. The gravity duals of such deformations have been identified by Lunin and Maldacena [3] and are constructed by applying an SL(3, R) transformation
or, equivalently [4], a certain sequence of T-dualities, S-dualities and coordinate shifts to the initial AdS5 S5 solution. This construction has been generalized in [46] and extended to other
backgrounds in [7] while diverse aspects of the deformation have been examined in [812].
The LuninMaldacena construction can be carried over to the Coulomb branch of the gauge
theory. The latter is obtained by moving away from the conformal point at the origin of moduli
space by giving nonzero vevs to the SO(6) scalars. The corresponding gravity duals are obtained by generalizing the stacked-brane distribution to a multicenter one, thereby breaking the
SO(6) isometry of the solutions [13]. A class of marginal deformations ( -deformations) of
these solutions have been obtained by the procedure outlined above in [14]. Probes of the resulting deformed geometries with Wilson loops, according to the recipe of [15], have revealed
a rich structure of phenomena in the gauge theory, with behaviors ranging from the standard
Coulombic interaction to complete screening and linear or logarithmic confinement, while the
wave equation for the radial modes of massless scalar excitations in the deformed backgrounds
turns out to be related to the Inozemtsev BC1 integrable system [14]. Another class of marginal
deformations ( -deformations) of these solutions has been obtained in [16], where Wilson-loop
calculations indicate the existence of a linear confining potential in some cases.
A further step forward would be to extend this construction to include the full set of complex
-deformations and to apply it for the most general case of non-extremal rotating D3-branes [13,
17,18] which are dual to the gauge theory at finite temperature and R-charge chemical potentials
and which include the multicenter D3-branes dual to the Coulomb branch as a limiting case.
The construction of these deformed type IIB backgrounds is the main purpose of this paper.
After constructing the deformed solutions, we explore diverse aspects of these backgrounds,
namely the Penrose limits of the multicenter solutions, the giant graviton states supported in the
resulting PP-waves and, finally, the Wilson-loop heavy-quark potential of the dual gauge theory.
These investigations are carried out with emphasis on -deformations, as their effect is often
overlooked in the literature although it is in many cases significant.
This paper is organized as follows: In Section 2 we present in detail the -deformation procedure for the most general type IIB background consisting of the metric, a 4-form potential and a
dilaton and possessing at least a U (1)3 global symmetry. In Section 3 we consider a class of such
backgrounds, corresponding to rotating and multicenter D3-branes, we specialize to three simple cases for which we present the explicit form of the deformed metrics, and we demonstrate
the expected equivalence of the thermodynamics of the deformed and undeformed metrics. In
Section 4 we construct the PP-wave backgrounds arising as Penrose limits of the deformed multicenter solutions along a certain set of BPS and non-BPS geodesics. In Section 5, we consider
the simplest PP-wave background of this type and we investigate the effect of -deformations
on the energetics of giant gravitons supported by this geometry. In Section 6 we probe the deformed multicenter geometries by static Wilson loops for the case of -deformations. Finally, in
Section 7 we summarize and conclude. Our conventions for T- and S-duality are summarized in
Appendix A.
57
2. Marginal deformations of type IIB backgrounds with U (1)3 isometry

2.1. Marginally deformed N = 4 SYM and its gravity dual
On the gauge-theory side, our general setup refers to a class of exactly marginal deformations
of N = 4 super-YangMills theory, namely the LeighStrassler -deformations of [2]. In this
construction, one starts from the N = 4 theory with complexified gauge coupling
=
4i
YM
+ 2 ,
2
gYM
(2.1)
and applies a deformation that acts on the three complex chiral superfields i , i = 1, 2, 3, of the
theory by modifying the standard superpotential W = Tr(1 [2 , 3 ]) to

W = Tr ei 1 2 3 ei 1 3 2 ,
(2.2)
where is a complex phase. The latter is conveniently parametrized as = , where
and are real parameters with unit period; in the special cases = 0 or = 0, the deformation
is referred to as a -deformation or a -deformation respectively. The above deformation breaks
N = 4 supersymmetry down to N = 1 and the corresponding SO(6)R global R-symmetry group
down to its U (1)1 U (1)2 U (1)R Cartan subgroup where U (1)R stands for the surviving Rsymmetry. There is also a Z3 symmetry under cyclic permutations. Under U (1)1 U (1)2 , the
charges of the chiral superfields are
1 2 3
1 2 3
Q2 , Q2 , Q2 = (1, 1, 0),
Q1 , Q1 , Q1 = (0, 1, 1),
(2.3)
while the superpotential is invariant. The Coulomb branch of the theory is described by the Fterm conditions
q e2i ,
1 2 = q2 1 ,
and cyclic,
(2.4)
which are valid for large N (exact for U (N )). For generic , these conditions are solved by traceless N N matrices, where in each entry at most one of them is nonzero. For -deformations
with rational , the Coulomb branch contains additional regions [10,11].
On the gravity side, the dual to the LeighStrassler deformation was constructed by Lunin
and Maldacena in [3] for field theories possessing at least a U (1)1 U (1)2 global symmetry
which, in the gravity dual, corresponds to an isometry of the supergravity background along two
angular directions parametrizing a 2-torus. The deformation proceeds by applying a certain dependent SL(3, R) transformation belonging to the first factor of the full SL(3, R) SL(2, R)
duality group of type IIB supergravity on the 2-torus. The effect of this transformation in the field
theory amounts to the replacement of the standard product of field operators by the Moyal-like
product
f
f g = ei(Q1 Q2 Q1 Q2 ) f g,
(2.5)
which, for the case of N = 4 SYM, does indeed induce the modified superpotential (2.2). The
transformation outlined above was applied to various type IIB backgrounds in [3] and further
generalized to a broader class of backgrounds in [7].
To illustrate the construction for the solutions of interest, we consider a general type IIB background where the ten spacetime coordinates are split into a seven-dimensional part parametrized
by x I , I = 1, 2, . . . , 7, and a three-dimensional part corresponding to a 3-torus parametrized by
58
the angles i , i = 1, 2, 3. The most general metric of this form is given by

2
ds10
= GI J (x) dx I dx J + 2
3
I i (x) dx I di +
i=1
3
zi (x) di2 ,
(2.6)
i=1
where the zi are positive-definite functions. This metric contains the mixed dx I di -terms and
generalizes that considered in [14] as the staring point for constructing marginally deformed
backgrounds. Apart from the metric, the solution is characterized by the dilaton and the RR
4-form
1
A4 = C4 + C3i di + ij k C2i dj dk + C1 d1 d2 d3 ,
2
(2.7)
where C1 , C2i , C3i and C4 are forms of degree indicated by the lower index. They have dependence and support only on x I and are constrained by the self-duality relations
dC1 d1 d2 d3 = dC4 ,

1
ij k dC2i dj dk = dC3i di .
2
(2.8)
A special case of the above solution, which includes the rotating branes to be examined in Section 3, is the one where the only nonzero I i are those in the ti directions, i ti . For this case,
the metric (2.6) simplifies to
2
= GI J (x) dx I dx J + 2
ds10
3

i=1
i (x) dt di +
3
zi (x) di2 .
(2.9)
i=1
According to the gauge/gravity correspondence, the three chiral superfields i are in oneto-one correspondence with three complex coordinates wi = Ri (x)eii and the generators
(Q1 , Q2 , QR ) of the global symmetries of the gauge theory correspond to linear combinations
of the generators (J1 , J2 , J3 ) of the shifts along the torus. To proceed, it is most convenient to
trade the i for a new set of variables i such that the generators (J1 , J2 , J3 ) of shifts along
these directions become precisely identified with (Q1 , Q2 , QR ). From (2.3), it is easily seen that
the appropriate set of variables is
1
1 = (1 + 2 23 ),
3
1
3 = (1 + 2 + 3 ).
3
1
2 = (2 + 3 21 ),
3
(2.10)
In terms of these new variables, the description of the SL(3, R) transformations corresponding
to -deformations in the gravity dual is rather simple. These transformations can be written [4]
as the sequence ST sT S 1 where T stands for a T-duality along the isometry direction 1 , S and
S 1 stand for an S-duality with parameter / and respectively (see Appendix A for
conventions), and s denotes the coordinate shift
s : 2 2 + 1 .
(2.11)
The special case of -deformations corresponds to the sequence T sT and is obtained from the
above case by taking = 0. The detailed procedure is presented below.
59
2.2. Construction of the marginally-deformed solutions

We now construct the -deformations of the gravity duals under consideration by applying
the ST sT S 1 sequence of transformations to the solution (2.6)(2.8). Here, we will denote the
metric components in the and bases by GMN and gMN respectively, while we will indicate
(1)
(2)
the various fields at the intermediate steps by superscripts, e.g., gMN , gMN , . . . , and those at the
final step by a hat. In the various computations we use the T- and S-duality rules that we have
written in Appendix A in a compact way particularly useful for our purposes. To present the
results in a succinct form, it is convenient to introduce the 2-forms
B2 = I 1 J 2 dx I dx J + z1 (I 2 I 3 ) d1 dx I + z1 z2 d1 d2 + cyclic,
A2 = C1 (d1 + d2 + d3 ) + C21 + C22 + C23 ,
(2.12)
the first of which can be written as the product of two one-forms B2 = A1 B1 with
A1 = (I 2 I 3 ) dx I + z2 d2 z3 d3 ,

I 2 z3 + I 3 z2
z2 z3
B1 =
I 1 dx I z1 d1 +
(d2 + d3 ),
z2 + z3
z2 + z3
(2.13)
and hence is nilpotent in the sense that

B2 B2 = 0.
(2.14)
It is also useful to introduce

G 1 = 1 + ||2 (z1 z2 + z2 z3 + z3 z1 ),
H = 1 + 2 e2 (z1 z2 + z2 z3 + z3 z1 ),
Q = e (z1 z2 + z2 z3 + z3 z1 ),
(2.15)
where
= = i e ,
(2.16)
since the axion in the initial solution is zero. The calculation proceeds as follows:
In the first step we perform an S-duality with parameter = / . Applying the transformation and passing to the basis, we obtain the metric
||
GI J ,
||
||
||
(1)
(1)
(1)
gI 2 =
gI 3 =
(I 2 I 3 ),
(I 2 I 1 ),
(I 1 + I 2 + I 3 ),
gI 1 =
||
||
||
(1)
(1)
(1)
g22 =
g33 =
(z2 + z3 ),
(z1 + z2 ),
(z1 + z2 + z3 ),
g11 =
||
||
||
(1)
(1)
(1)
g13 =
g23 =
z2 ,
(z2 z3 ),
(z2 z1 ),
g12 =
(2.17)
gI(1)
J =
the dilaton
e2
(1)
||4 2
e ,
4
(2.18)
60
and the axion

(1)
A0 =
2
e
.
||2
(2.19)
The RR 4-form remains unchanged and, in the new basis, reads

(1)
A4 = A2 d1 d2 + C21 + C22 2C23 d2 d3

+ C22 + C23 2C21 d3 d1

+ C32 C33 d1 + C32 C31 d2 + C31 + C32 + C33 d3 + C4 ,
(2.20)
while the NSNS and RR 2-forms are zero.

In the second step we perform a T-duality along 1 . The NSNS fields are given by

||
(I 2 I 3 )(J 2 J 3 )
(2)
,
GI J
gI J =
z2 + z3

|| z2 I 3 + z3 I 2
|| 2(z2 I 3 + z3 I 2 )
(2)
(2)
gI 2 =
gI 3 =
I 1 ,
+ I 1 ,
z2 + z3
z2 + z3

1
|| z1 z2 + z2 z3 + z3 z1
||
4z2 z3
(2)
(2)
(2)
g11 =
,
z1 +
,
g22 =
,
g33 =
|| z2 + z3
z2 + z3
z2 + z3

|| 2z2 z3
(2)
g23
=
z1 ,
z2 + z3

1
(2)
B2 =
(I 2 I 3 ) dx I + z2 d2 + (z2 z3 ) d3 d1 ,
z2 + z3
||3 e2
(2)
e2 = 3
(2.21)
,
z2 + z3
while the RR fields are
(2)
A1 = 2 e2 d1 ,
||

(2)
A3 = A2 d2 + C22 + C23 2C21 d3 + C32 C33 ,
(2)
(1)
(2)
(2)
A5 = A4 d1 + B2 A3 .
(2.22)
In the third step we perform a coordinate shift 2 2 + 1 . The changed metric components are

z2 I 3 + z3 I 2
(3)
I 1 ,
gI 1 = ||
z2 + z3
1 1
(3)
G
g11 =
,
||
z2 + z3

z1 z2 + z2 z3 + z3 z1
2z2 z3
(3)
(3)
,
g13 = ||
z1 .
g12 = ||
(2.23)
z2 + z3
z2 + z3
The 3-form changes as
(3)
(2)
A3 = A3 A2 d1 ,
while all other fields remain invariant.
(2.24)
61
In the fourth step another T-duality along 1 is performed. Applying the duality and returning to the basis, we find that the NSNS fields are given by

||
GI J ||2 G z1 (I 2 I 3 )(J 2 J 3 ) + cyclic ,

||
(4)
G I i + ||2 (z1 z2 I 3 + cyclic) ,
GI i =

||
(4)
Gij =
G zi ij + ||2 z1 z2 z3 ,
||2
(4)
B2 =
GB2 ,
||4
(4)
e2 = 4 Ge2 ,
(4)
GI J =
while the RR fields are
(4)
A0 = 2 e2 ,
||
(2.25)
A2 = A2 e2 GB2 ,
(4)
2
A(4)
4 = A4 || GB2 A2 ,
A(4)
6 =
||2
GA4 B2 .
(2.26)
The final step is another S-duality, now with parameter . This leads to the gravity dual
of the -deformed theory, expressed in terms of the NSNS fields

I J = H1/2 GI J ||2 G z1 (I 2 I 3 )(J 2 J 3 ) + cyclic ,

G

I i = GH1/2 I i + ||2 (z1 z2 I 3 + cyclic) ,
G

ij = GH1/2 zi ij + ||2 z1 z2 z3 ,
G
B 2 = GB2 A2 ,
e2 = GH2 e2 ,
(2.27)
and the RR fields

A 0 = H1 Qe ,
A 2 = A2 e2 GB2 ,
1
A 4 = A4 2 GB2 A2 + A2 A2 ,
2
A 6 = B 2 A 4 ,
(2.28)
where in the last relation we made use of the nilpotency of B2 (2.14). One can check that the
above formulas reduce, for the appropriate limiting cases, to the various solutions that have been
found in the literature [3,4,14]. The case of -deformations is obtained as the special case of the
above where = 0, in which case the quantities in Eq. (2.15) reduce to
G 1 = 1 + 2 (z1 z2 + z2 z3 + z3 z1 ),
H = 1,
Q = 0.
(2.29)
At this point, it is instructive to compare the general case of -deformations with the special
case of -deformations. We see that the extra -dependence in the former enters through (i) the
replacement 2 ||2 in the metric and in the definition of G, (ii) the overall factors H1/2
62
and H2 in the metric and dilaton respectively and (iii) a nonzero axion as well as new terms
in the NSNS 2-form, the RR 2-form and the RR 4-form proportional to A2 , B2 and A2 A2 ,
respectively. These changes clearly affect the NambuGoto (or DiracBornInfeld plus Wess
Zumino) actions of probe strings (or branes) propagating in the deformed geometry and so, in
relevant investigations, one is entitled to expect qualitative departures from results obtained for
purely -deformed
backgrounds. On the other hand, one readily verifies that the massless scalar
wave equation M ( Ge2 GMN N ) = 0 is insensitive to the presence of the H factors,
which implies that its analysis proceeds as in the -deformed case with the replacement 2
||2 and that, in the case of multicenter D3-branes considered in [14], its relation with integrable
systems found there remains intact.
3. Deformations of rotating and multicenter D3-branes
In this section, we explicitly apply the -deformation procedure described above to rotating
D3-brane solutions. First, we give a brief review of the field-theory limit of these solutions and
we focus on three special cases where the metrics simplify considerably. Then, we present the
full metrics of the corresponding deformed solutions. Finally, we demonstrate that the thermodynamic properties of the deformed metrics are exactly the same as for the undeformed ones.
3.1. Rotating and multicenter D3-brane solutions
The solutions we are interested in here are the non-extremal rotating D3-branes found in full
generality in [13] using previous results from [17]. They are characterized by the non-extremality
parameter plus the rotation parameters ai , i = 1, 2, 3, which correspond to the three generators
in the Cartan subalgebra of SO(6). The spacetime coordinates are split into the brane coordinates
(t, x3 ) = (t, x1 , x2 , x3 ) and the transverse coordinates ym , m = 1, . . . , 6, which can be parametrized as

w1 = y1 + iy2 = r 2 + a12 sin ei1 ,

w2 = y3 + iy4 = r 2 + a22 cos sin ei2 ,

w3 = y5 + iy6 = r 2 + a32 cos cos ei3 ,
(3.1)
where the complex coordinates wi are in one-to-one correspondence with the chiral superfields i of the gauge theory. Here we are interested in the field-theory limit of these solutions
which, in the most general non-extremal case, is given by the metric (we follow [18], in which
the thermodynamic properties of the solution, presented below in some special cases, were also
computed)

r 6 2
4
ds 2 = H 1/2 1 4
dr
dt 2 + d x32 + H 1/2
f
r

+ H 1/2 r 2 1 d 2 + r 2 2 cos2 d 2 + 2 a22 a32 cos sin cos sin d d

+ r 2 + a12 sin2 d12 + r 2 + a22 cos2 sin2 d22 + r 2 + a32 cos2 cos2 d32

2
2
2
2
2
2
2 2 dt a1 sin d1 + a2 cos sin d2 + a3 cos cos d3 ,
(3.2)
R
63
the constant dilaton

e = e0 = gs ,
(3.3)
and a 4-form potential of the form (2.8) with

H 1
dt dx1 dx2 dx3 ,
gs

1 2 3
2
C3 , C3 , C3 = 2 a1 sin2 , a2 cos2 sin2 , a3 cos2 cos2 dx1 dx2 dx3 ,
R gs
(3.4)
C4 =
and C1 and C2i specified by the duality relations (2.8). In the above, the various functions are
given by
R4
,
r 4

2
f = r + a12 r 2 + a22 r 2 + a32 4 r 2 ,
H=
a32 2

a22 2
a12
2
2
2
sin
cos
sin
+
cos
+ 2 sin cos2 + sin2

2
2
r
r
r
2a2
a12 a32
a22 a32 2
a
+ 4 sin + 4 cos2 sin2 + 1 4 2 cos2 cos2 ,
r
r
r
2
2
2
a
a
a
1 = 1 + 12 cos2 + 22 sin2 sin2 + 32 sin2 cos2 ,
r
r
r
2
2
a
a
2 = 1 + 22 cos2 + 32 sin2 .
r
r
In the notation of Section 2, the metric components zi and i are given by
=1+
(3.5)

R 2 2
r + a12 sin2 , r 2 + a22 cos2 sin2 , r 2 + a32 cos2 cos2 ,
r 2 1/2

2
(1 , 2 , 3 ) = 2 1/2 a1 sin2 , a2 cos2 sin2 , a3 cos2 cos2 .
(3.6)
r
For = 0, these solutions describe rotating branes and are dual to N = 4 SYM at finite temperature and R-charge chemical potentials, and the ai parametrize the angular velocities/momenta
on the supergravity side and the chemical potentials/R-charges on the gauge theory-side. For
= 0, these solutions describe multicenter brane distributions dual to the Coulomb branch of
N = 4 SYM, and the ai parametrize the principal radii of the distribution on the supergravity
side and the scalar vevs on the gauge-theory side. In the rest of the paper, we will refer to the ai
as rotation parameters, keeping in mind their different interpretations in these two cases.
In what follows, we examine some simple special cases of the above general solution, namely
those corresponding to three equal nonzero rotation parameters, two equal nonzero rotation parameters and one nonzero rotation parameter.
(z1 , z2 , z3 ) =
3.1.1. Three equal rotation parameters

The first special case we consider is the one where all three rotation parameters are equal to
each other, a1 = a2 = a3 = r0 . Employing the change of variable r 2 r 2 r02 we write the
64
resulting metric as

4
r6
2
2
1 4 dt + d x3 + H 1/2
dr 2
ds = H
r
r 6 4 (r 2 r02 )

2
22 r0 2
2
2
,
dt
sin
d
+
cos
sin
d
+
cos
d
+ H 1/2 r 2 d52
1
2
3
R2
(3.7)
where
1/2
H=
R4
,
r4
(3.8)
and d52 is the S5 metric

d52 = d 2 + sin2 d12 + cos2 d 2 + sin2 d22 + cos2 d32 .
(3.9)
The horizon radius rH for this metric is given by the largest root of the equation

r 6 4 r 2 r02 = 0,
(3.10)
while its Hawking temperature reads

T=
2 3r 2 )
rH (2rH
0
2 r 2)
2R 2 (rH
0
(3.11)
For = 0 this background reduces to the AdS5 S5 background obtained by stacked D3-branes
at the origin.
3.1.2. Two equal rotation parameters
The second special case is the one where two rotation parameters are set to the same nonzero
value, which we may take as a2 = a3 = r0 . Employing again the change of variable r 2 r 2 r02
we have the metric

r 4 (r 2 r02 cos2 )
4 H
2
1/2
2
2
1 4 dt + d x3 + H 1/2
ds = H
dr 2
R
(r 4 4 )(r 2 r02 )

+ H 1/2 r 2 r02 cos2 d 2 + r 2 cos2 d32 + r 2 r02 sin2 d12

2
2 r0
2
2
2 2 dt cos sin d2 + cos d3 ,
R
(3.12)
where
H=
R4
,
r 2 (r 2 r02 cos2 )
(3.13)
while d32 is the S3 metric

d32 = d 2 + sin2 d22 + cos2 d32 .
(3.14)
Now, the horizon radius is simply

rH = ,
(3.15)
65
and the Hawking temperature reads

2 r02
T=
(3.16)
.
R 2
For = 0 this background reduces to that obtained by a uniform distribution of D3-branes on a
3-sphere of radius r0 .
3.1.3. One rotation parameter
The third special case is the one where there is only one nonzero rotation parameter, which
we may take as a1 = r0 . In this case, we have the metric

r 2 (r 2 + r02 cos2 ) 2
4 H
ds 2 = H 1/2 1 4 dt 2 + d x32 + H 1/2
dr
R
r 4 + r02 r 2 4

2
1/2
+H
r + r02 cos2 d 2 + r 2 cos2 d32 + r 2 + r02 sin2 d12

2 r0 2
sin
dt
d
1 ,
R2
where the harmonic function is given by
2
H=
(3.17)
R4
,
r 2 (r 2 + r02 cos2 )
(3.18)
while d32 is defined as before. The horizon radius is given by

1
2
rH
(3.19)
= r02 + r04 + 44 ,
2
and the Hawking temperature reads

rH r04 + 44
T=
(3.20)
.
2R 2 2
For = 0 this background reduces to that obtained by a uniform distribution of D3-branes on a
disc of radius r0 and is related to the corresponding background for two rotation parameters by
the transformation r02 r02 .
3.2. The deformed metrics
After the above preliminaries, we are ready to apply the marginal-deformation procedure to
the rotating-brane solutions just discussed. In what follows, we present the explicit form of the
marginally-deformed metrics for the three special cases considered earlier. To keep the notation
as compact as possible, it is convenient to introduce the shorthand notation c cos and s
sin and the following rescaling for the deformation parameters1
R2

,
2
R2
,
2
R2
,
2gs
2 = 2 + 2 .
with the new parameters satisfying ||
1 These parameters differ by the analogous parameters in [14] by a factor of 1/2.
(3.21)
66
3.2.1. The zero-rotation case

As a first example, let us consider the case where all rotation parameters are set to zero. Setting
r0 = 0 in any of the above three metrics and substituting in (2.26), we find

4
2
1/2
1/2
2
2
ds = H
H
1 4 dt + d x3
r

4
r
1/2
2
2
2
+H
(3.22)
dr + r d5, ,
r 4 4
2 is the metric on the deformed five-sphere S5 , given by
where d5,
2

2
2
2
2
2
2
2
d5, = d + Gs d1 + c d + G s d2 + c2 d32
3
2

2
4
2
2
c s s2
di ,
+ G||
(3.23)
i=1
and the various functions are

2 c2 s2 + c2 c2 s2 ,
G 1 = 1 + 4||

H = 1 + 4 2 c2 s2 + c2 c2 s2 ,
(3.24)
and H = R 4 /r 4 . The resulting space is thus a conformal rescaling of the product of AdS5 Schwarzschild with S5 .
3.2.2. Three equal rotation parameters
Starting from the case with three equal rotation parameters, the deformed metric is found by
substituting the undeformed metric (3.7) into Eq. (2.26). After some algebra, we find

2 Gr 2 c2 (c4 s2 + 4s4 + 4c2 s2 c4 )]
4 [r 2 ||

0 2
2
2
ds 2 = H1/2 H 1/2 1
+
d
x

dt
3
r6

r6
2
dr 2 + r 2 d5,
+ H1/2 H 1/2
r 6 4 (r 2 r02 )

3
2

1/2 1/2 2 r0
2
2 2
2
2
4
2
2
c s s2
dt s d1 + c s d2 + c d3 + 3||
di ,
GH H
R2
i=1
(3.25)
2 is as in (3.23), G 1 and H are as in (3.24) and H is as in (3.8).
where d5,
3.2.3. Two equal rotation parameters

For the case with two equal rotation parameters, the deformed metric is found to be

4 [1
2
1/2 1/2
ds = H H
1

+H
2 G r 2 c4 (4(r 2 r 2 )s2 c2 +r 2 c2 s2 )
||
2
0
0 2
r 4 (r 2 r02 c2 )
r 2 (r 2 r02 c2 )
2
2
r (r r02 c2 )
1/2

dt 2 + d x32

dr + r
d
(r 2 r02 )(r 4 4 )

2
1/2 1/2
+ GH H
r r02 s2 d12 + r 2 c2 s2 d22 + c2 d32
1/2
r02 c2
+ r 2 c2 d 2
2 r 2 (r 2 r 2 )c4 s2 s2
||
0 2
67
2
3
di
r 2 r02 c2
i=1

3
2 (r 2 r 2 )c4 s2 s2

2||
22 r0
0 2
2 2
2
dt c s d2 + c d3 +
di ,
R2
r 2 r02 c2
(3.26)
i=1
with
G
2 c2
= 1 + 4||
H = 1 + 4 2 c2
(r 2 r02 )s2 + r 2 c2 c2 s2
r 2 r02 c2
(r 2 r02 )s2 + r 2 c2 c2 s2
r 2 r02 c2
,
(3.27)
and H as in (3.13).
3.2.4. One rotation parameter
For the case with one rotation parameter, the deformed metric is found to be

2 Gr 2 c2 s4 )
4 (r 2 + r02 c2 4||
0
2
1/2 1/2
2
2
1
dt + d x3
ds = H H
r 2 (r 2 + r02 c2 )2
2 2

2 2
2
2
1/2 1/2 r (r + r0 c )
2
2 2
2 2
2
+H H
dr + r + r0 c d + r c d
r 4 + r02 r 2 4

2
1/2 1/2
+ GH H
r + r02 s2 d12 + r 2 c2 s2 d22 + c2 d32
+
2 r 2 (r 2 + r 2 )c4 s2 s2
||
0 2
r 2 + r02 c2

3
2
di
i=1
3
2 r 2 c4 s2 s2
||
22 r0
2
2
dt
s
d
+
di
1
R2
r 2 + r02 c2 i=1

,
(3.28)
with
2 c2
G 1 = 1 + 4||
H = 1 + 4 2 c2
(r 2 + r02 )s2 + r 2 c2 c2 s2
r 2 + r02 c2
(r 2 + r02 )s2 + r 2 c2 c2 s2
r 2 + r02 c2
,
(3.29)
and H as in (3.18).
3.3. Invariance of the thermodynamics
A useful consistency check for our calculation is to examine the thermodynamic properties of
the deformed rotating-brane solutions. In particular, since the deformed solutions are related to
the undeformed ones by U-duality transformations that are symmetries of the underlying theory,
all thermodynamic quantities for the deformed metrics must be equal to those for the undeformed
68
ones. It is instructive to show that this is indeed the case by explicitly calculating the angular
velocities, the Hawking temperature, and the entropy for the general deformed solutions.
We start by writing the deformation (2.27), restricted to a metric of the form (2.9), in the Ein
(E) = e/2
MN = gs1/2 G 1/4 H1/2 G
MN,
stein frame. In this frame, the deformed metric is G
G
MN
and thus we have

(E) = gs1/2 G 1/4 GI J ||2 G z1 (2 3 )2 + cyclic I,t J,t ,
G
IJ

(E) = gs1/2 G 3/4 i + ||2 (z1 z2 3 + cyclic) ,
G
ti

(E) = gs1/2 G 3/4 zi ij + ||2 z1 z2 z3 ,
G
(3.30)
ij
where we recall that, for the general rotating-brane solution (3.2), zi and i are given in (3.6)
1/2
(E)
and G and H are given in terms of the zi in (2.15). Letting GMN = e0 /2 GMN = gs
GMN
be the undeformed metric in the Einstein frame, we write
(E) = G(E) + G(E) ,
G
MN
MN
MN
(3.31)
(E)
where the functions GMN represent the effect of the deformation and are read off from (3.30)
and the explicit relations (3.6) for zi and i and (2.15) for G and H. Although the resulting
(E)
GMN are very complicated functions of r, and , inspection of (2.15) and (3.6) shows that
they satisfy
(E)
(E)
(E)
GMN (,)=(/2,0) = GMN (,)=(0,/2) = GMN (,)=(0,0) = 0,
(3.32)
while it can be shown that
(E)
r,, G
MN (,)=(/2,0)
(E)
= r,, GMN (,)=(0,/2)
(E)
= r,, GMN (,)=(0,0) = 0.
(3.33)
That is, there exist three values of (, ) for which the metric and its derivatives reduce to those
in the undeformed case.
Given Eqs. (3.32) and (3.33), it is very easy to check that the angular velocities and the Hawking temperature are the same as in the undeformed solution. Indeed, it immediately follows that
the horizon radius for the deformed metric is equal to the horizon radius rH for the undeformed
one. The angular velocities i , are found by demanding that the Killing vector = t + i i
associated with a stationary observer be null at r = rH , i.e., by solving the equation
(E)
(E)

2 (rH ) = 2 (rH ) + G(E)
tt (rH ) + 2Gti (rH )i + Gij (rH )i j = 0,
(3.34)
(E)
where 2 and 2 are the norms of with the metrics G(E)
MN and GMN , respectively. Evaluating this
equation at (, ) = (/2, 0), (0, /2) and (0, 0) and using (3.32), we obtain three decoupled
equations for 1 , 2 and 3 , respectively, which are the same ones that arise for the undeformed
metric [18]. Therefore, the angular velocities are the same, i = i . The Hawking temperature
is found from the relation
(E)MN M 2 N 2
1
G
TH2 =
(3.35)
lim
,
16 2 rrH
2
which, being independent of the angles, can be evaluated at any of the aforementioned three
values of (, ). Then, use of (3.32) and (3.33) leads to the same relation as for the undeformed
metric and so TH = TH .
69
Finally, the entropy is determined by the horizon area which is in turn related to the determinant of the eight-dimensional metric along the directions normal to the horizon. Labelling
these directions as A = (, i) with = (
x3 , , ) and i = (1 , 2 , 3 ), we find that this eightdimensional metric equals

0
1/2 G
(E)
GAB = gs
(3.36)
0
zi ij
and
(E) = gs1/2
G
AB
G 1/4 G
0

0
,
G 3/4 (zi ij + ||2 z1 z2 z3 )
(3.37)
for the undeformed and deformed cases, respectively. Then, a simple calculation gives

(E) = gs4 G 1 + ||2 (z1 z2 + z2 z3 + z3 z1 ) z1 z2 z3 det G
det G
AB
= gs4 z1 z2 z3 det G = det GAB ,
(E)
(3.38)
where in the second line we used the defining relation for G. Therefore the entropy is indeed
invariant, a fact that seems to be closely related to the invariance of the central charge of the dual
CFT under deformations [9]. This completes our consistency check.
4. Penrose limits of the -deformed solutions
Having constructed the deformed solutions, it is interesting to examine their Penrose limits,
following by applying the standard procedure introduced by [19] to the metric and the other
fields of the solution.2 The resulting spacetimes are PP-waves that constitute generalizations of
the maximally supersymmetric PP-wave solution [22] of type IIB string theory and, in the context
of the AdS/CFT correspondence, are related to the BMN limit [23,24] of the gauge theory. PPwave limits of marginally-deformed backgrounds were first considered in [25] and in [3], with
the former construction starting from the gauge-theory side of the correspondence, while further
aspects of such solutions were examined in [26]. Here, we further generalize these constructions
to include the effects of -deformations and of turning on rotation parameters by following [27]
where PP-wave solutions based on the solutions of Section 3.1 were constructed and further
analyzed.
4.1. Null geodesics in the deformed geometry
To find the Penrose limits of the deformed solutions, we first need to identify null geodesics
in the respective geometries. The geodesics we are interested in involve t , r and one linear combination of the cyclic coordinates i which we denote by , taking the remaining coordinates to
constant values consistent with their equations of motion. To seek such geodesics, we note that
the i are cyclic and hence setting any of them to any constant value is automatically consistent,
while an ansatz with constant and is certainly consistent if Gij i j = Gij i j = 0 and
Gti i = Gti i = 0, where the dot denotes differentiation with respect to the affine parameter of the geodesic. In our examples with nonzero rotation (r0 = 0), these equations are solved
2 In a string theory context, in the presence of non-vanishing scalar and tensor fields, the Penrose limiting procedure
for constructing PP-wave solutions was first applied in [20,21].
70
if
=
,
2
= 0,
, i = any,
= 0,
,
2
i = any,
1 = any.
,
2 = 3 ,
4
For the case of zero rotation (r0 = 0), there emerge additional solutions, one of which is
= 0,
1
= arcsin ,
3
,
4
1 = 2 = 3 .
(4.1)
(4.2)
Regarding the sensitivity of the metrics along these geodesics to -deformations, we note that
the latter are non-trivial only when z1 z2 + z2 z3 + z3 z1 = 0 which, by (3.6), requires that = /2
and (, ) = (0, 0), (0, /2). Therefore, the effective metrics along the geodesics in the first two
lines of (4.1) are insensitive to -deformations while those along the geodesics in the third line
of (4.1) and in (4.2) are sensitive to -deformations. Setting the unspecified i to zero, we are
led to consider the following cases
(J, 0, 0): = , = 0, 1 = , 2 = 3 = 0,
2
(0, J, 0): = 0, = , 2 = , 3 = 1 = 0,
2
(0, 0, J ): = 0, = 0, 3 = , 1 = 2 = 0,
1
(J, J, J ): = arcsin , = , 3 = , 1 = 2 = 0, for r0 = 0,
4
3
2 + 3
2 3
= ,
= 0, 1 = 0, (4.3)
(0, J, J ): = 0, = ,
4
2
2
where, in the fourth line, the i are as given in (2.10). The above cases correspond to a particle moving with angular momenta (J1 , J2 , J3 ) = (J, 0, 0), (0, J, 0), (0, 0, J ), (J, J, J ) and
(0, J, J ) respectively along the three isometry directions.
To examine the properties of these geodesics in the various backgrounds, we distinguish the
following cases:
Undeformed, zero rotation. In this case, the five-sphere is round and the full SO(6) isometry
group operates. All choices correspond to BPS geodesics that can be rotated into one another.
Deformed, zero rotation. Here, the five-sphere is deformed, with the isometry group broken
to U (1)3 . The choices (J, 0, 0), (0, J, 0) and (0, 0, J ) correspond to three BPS geodesics
that can still be rotated into each other, the choice (J, J, J ) corresponds to a distinct BPS
geodesic, and the choice (0, J, J ) corresponds to a distinct non-BPS geodesic.
Deformed, nonzero rotation. Now, the deformed five-sphere is in addition squashed. The
available choices (J, 0, 0), (0, J, 0), (0, 0, J ) and (0, J, J ) are all generically inequivalent
but, for the specific backgrounds considered in Section 3.2, the choices (0, J, 0) and (0, 0, J )
remain equivalent and it suffices to consider only one of them, say the second.
We note that the (J, J, J ) geodesic has been first considered for undeformed AdS5 S5 in
[25] (see also [26]), which is also where Penrose limits of marginally-deformed AdS5 S5 first
appeared, found through field-theory considerations.
71
4.2. The Penrose limit

Having identified the geodesics of interest, we are ready to take the Penrose limit, proceeding along the lines of [27]. We first employ the rescaling (t, x3 ) R 2 (t, x3 ) and we write the
effective metric for t , r and as
ds32
= tt dt 2 + rr dr 2 + d 2 + 2t dt d.
(4.4)
R2
Independence of the metric from t and leads to two conserved quantities, associated with the
Killing vectors k = t and l = and identified with the energy and the angular momentum,
namely
E = k u = tt t t = 1,
J = l u = t t .
(4.5)
Solving for t and and substituting into ds32 = 0, we obtain the equation
r 2 =
+ 2J t + J 2 tt
2 )
rr (t
tt
(4.6)
whose solution determines r in terms of . We next change variables from (r, t, ) to the new
variables (u, v, y), defined according to

+ 2J t + J 2 tt
dr =
du,
2 )
rr (t
tt
dt =
+ J t
2
t
d =
tt
du
t + J tt
2
t
tt
1
J
dv + dx,
2
R
R
du +
1
dx,
R
(4.7)
so that, in particular, u is identified with the affine parameter . We then rescale the spatial brane
coordinates as
r3
x3 = .
(4.8)
R
Finally, depending on the case at hand, we make the following changes of angular variables
(J, 0, 0): = ,
2
R
1
2
(0, 0, J ): = ,
= ,
R
R
y1
1
2y3
y3 3y4
y2
(J, J, J ): = arcsin + , = + , 1 =
,
, 2 =
R
4
R
R
3
2R
1
(0, J, J ): = ,
(4.9)
= + ,
= .
R
4
R
R
The Penrose limit is then obtained by substituting all these changes of variables into the original
solution and taking the limit R . For the marginally-deformed metrics of interest, this limit
must be taken while keeping R 2 and R 2 fixed. The resulting metric always takes the
form of a PP-wave in Rosen-like coordinates and can be brought to Brinkmann-like coordinates
by suitable changes of variables.
72
4.3. PP-wave limits of the deformed solutions

Here, we employ the method described above to derive the PP-wave limit of the marginallydeformed metrics of Section 3.2. To keep things relatively simple, we consider only the case
= 0, corresponding to D3-brane distribution, in which case the rotation parameter r0 is to
be thought of as the radius of the D3-brane distribution. Note also that, although the (J, 0, 0)
and (0, 0, J ) geodesics are insensitive to the -deformation, the PP-wave backgrounds resulting
from the limiting procedure described above are sensitive to the deformation.
4.3.1. The zero-rotation case
We begin with the zero-rotation case where the deformed metric is a conformal rescaling of
AdS5 S5 . The results for the various geodesics are as follows:
(J, 0, 0) geodesic. As a warmup exercise, and for later reference, we first review the construction of the simplest PP-wave limit of marginally-deformed AdS5 S5 , first derived in [3],
in some detail. The differential equation for r becomes
r 2 = 1 J 2 r 2 ,
(4.10)
with solution
r 2 (u) =
1
sin2 J u.
J2
(4.11)
After following the above limiting procedure we find that the Penrose limit of the deformed
metric reads
ds 2 = 2 du dv + Ar d r32 + Ax dx 2 + d y42 C du2 ,
(4.12)
where y4 is defined by

d y42 = d 2 + 2 d 2 + sin2 d22 + cos2 d32 ,
(4.13)
and the various functions are given by

1
Ar = r 2 = 2 sin2 J u,
J

2 J 2 y42 .
C = 1 + 4||
Ax = 1 J 2 r 2 = cos2 J u,
(4.14)
For future reference, we also write the Rosen form of the remaining nonzero fields, although we
will not do so for the remaining cases considered below. We have

B2 = 2J 2 du sin2 d2 cos2 d3 ,
e2 = gs2 ,

2J 2
du sin2 d2 cos2 d3 ,
gs
4

J sin J u
4
A4 =
dr
dr
dr
dx
+
cos
sin
du
d
1
2
3
2
3 . (4.15)
gs
J4
A2 =
73
Applying standard transformations3 to pass to Brinkmann coordinates, we write the metric as

ds 2 = 2 du dv + d r42 + d y42 + Fr r42 + Fy y42 du2 ,
(4.16)
where r4 = (r3 , x) and
Fr = J 2 ,

2 J 2,
Fy = 1 + 4||
(4.17)
and the remaining nonzero fields of the solution as

H3 = 4J du (dy1 dy2 dy3 dy4 ),
e2 = gs2 ,
4J
du (dy1 dy2 dy3 dy4 ),
F3 =
gs
J
F5 = du (dr1 dr2 dr3 dr4 dy1 dy2 dy3 dy4 ).
gs
(4.18)
We see that the deformation affects only the function Fy in the metric and the NSNS and RR
3-form field strengths.
(0, 0, J ) geodesic. The solution for r(u) is the same as before and the Penrose limit of the
metric in Rosen coordinates is given by (4.12) and (4.14) where y4 is now defined by

d y42 = dy12 + dy22 + dy32 + dy42 = d12 + 12 d12 + d22 + 22 d22 .
(4.19)
In Brinkmann coordinates, the Penrose limit is given by (4.17) and (4.18). We explicitly verify
that, for r0 = 0, the Penrose limits for the (J, 0, 0) and (0, J, 0) geodesics are equivalent as
already remarked in the comments following (4.3).
(J, J, J ) geodesic. For this case, it is convenient to introduce
2
4||
4 2
(4.20)
,
H=1+
,
3
3
which are just the effective constant values of the functions in (3.24) for the given ansatz. Then,
the differential equation for r has the form
G 1 = 1 +
Hr 2 = 1 J 2 r 2 ,
(4.21)
with solution

1
J
2
r (u) = 2 sin
u .
J
H1/2
2
(4.22)
In Rosen-like coordinates, the Penrose limit of the deformed metric reads

ds 2 = 2 du dv + Ar d r32 + Ax dx 2 + Ay d y22 + A y d y 22
+ By (y1 dy3 y2 dy4 ) du C du2 ,
(4.23)
3 For a metric in the Rosen form ds 2 = 2 du dv + A (u) dx 2 C du2 , applying the sequence of coordinate transfori i
i

d ln A
x
mations xi i and v v + 14 i du i xi2 brings it to the Brinkmann form ds 2 = 2 du dv + i dxi2 +( i Fi xi2
A
i
d 2 ln Ai
d ln A
C) du2 , where Fi = 14 ( du i )2 + 12
.
du2
74
where
d y22 = dy12 + dy22 ,
d y 22 = dy32 + dy42 ,
(4.24)
and

H1/2 2
J
Ar = H r = 2 sin
u ,
J
H1/2

J
u ,
Ax = 1 J 2 r 2 = cos2
H1/2
A y = GH1/2 ,
By = 4GJ,
Ay = H1/2 ,
2
2
2
G y
16J ||
2
.
C=
3H1/2
In Brinkmann-like coordinates, the metric reads
1/2 2
(4.25)
ds 2 = 2 du dv + d r42 + d y22 + d y 22 + G2 (y1 dy3 y2 dy4 ) du

+ Fr r42 + Fy y22 du2 ,
(4.26)

2G
J2
16J 2 ||
,
Fy =
,
H
3H
and the remaining fields of the solution are
Fr =
G2 =
4G 1/2
,
H1/2

4J
H3 = du 2 gs dy1 dy2 + G 1/2 (dy1 dy4 + dy2 dy3 ) ,
3H
2
e = GH2 gs2 ,

4J
F3 = du 2 dy1 dy2 G 1/2 (dy1 dy4 + dy2 dy3 ) ,

gs
3H
J
F5 = 3/2 du (dr1 dr2 dr3 dr4 dy1 dy2 dy3 dy4 ).
H gs
(4.27)
(4.28)
This is the generalization of the PP-wave considered in [25,26] which includes the effect of deformations. Now, the deformation affects all the F -functions in the metric and all nonzero
fields. This type of PP-wave falls into the subclass of homogeneous plane waves considered in
[28].
(0, J, J ) geodesic. For this case, we introduce
2,
G 1 = 1 + ||
H = 1 + 2 ,
(4.29)
which are again the effective constant values of the functions in (3.24). Then, the differential
equation for r has the form
Hr 2 = 1
with solution
r 2 (u) =
J 2r 2
,
G

G
J
2
sin
u
.
J2
G 1/2 H1/2
(4.30)
(4.31)
75
In Rosen-like coordinates, the Penrose limit of the deformed metric reads

ds 2 = 2 du dv + Ar d r32 + Ax dx 2 + Ay d y22 + By (y1 dy2 y2 dy1 ) du
+ A d 2 + A d 2 + B d du C du2 ,
(4.32)
where
d y22 = d12 + 12 d12 ,
(4.33)
and

J
GH1/2 2
Ar = H r =
sin
u ,
J2
G 1/2 H1/2

J
Ax = H1/2 G J 2 r 2 = GH1/2 cos2 1/2 1/2 u ,
G H
1/2
2
J,
Ay = H ,
By = 4||
1/2 2
A = H1/2 ,
C=
A = GH1/2 ,
B = 4J,
2 4||
4 )
22]
J 2 [(1 ||
y22 4||
.
H1/2
(4.34)
In Brinkmann-like coordinates, the metric is given by

ds 2 = 2 du dv + d r42 + d y22 + d 2 + d 2 + Gy (y1 dy2 y2 dy1 ) du

+ G d du + Fr r42 + Fy y22 + F 2 du2 ,
(4.35)

Fr =
J2
,
GH
2 4||
4)
2
J 2 (1 ||
4J ||
,
Gy =
,
H
H1/2
2
4J 2 ||
4J
,
G = 1/2 1/2 ,
F =
H
G H
Fy =
(4.36)
while the remaining fields read

e2 = GH2 gs2 ,
J
F5 = 1/2 3/2 du (dr1 dr2 dr3 dr4 dy1 dy2 d d ).
G H gs
(4.37)
We see that only the dilaton and the RR 5-form field strength survive the Penrose limit along this
particular geodesic. The deformation affects all the F -functions in the metric and all nonzero
fields.
4.3.2. Two equal rotation parameters (sphere)
We next consider the case of two equal rotation parameters, where the new parameter entering
the problem is r0 . The results for the various geodesics are as follows:
76
(J, 0, 0) geodesic. The differential equation for r becomes

r2
r 2 = 2 J 2 r 2 , 1 02 .
r
Its solution is then given by
1
(1 a cos 2J u), a
2J 2
from which it follows that

a2 1
(u) = 1 +
.
2(1 a cos 2J u)
r 2 (u) =
1 4J 2 r02 ,
(4.38)
(4.39)
(4.40)
Note that reality requires that, for fixed r0 , there is a maximum angular momentum associated
with this trajectory. In Rosen coordinates, the Penrose limit of the deformed metric reads
ds 2 = 2 du dv + Ar d r32 + Ax dx 2 + d y42 C du2 ,
(4.41)
where

d y42 = d 2 + 2 d 2 + sin2 d22 + cos2 d32 ,
(4.42)
and
Ar = r 2 ,
Ax = 2 J 2 r 2 ,

2 J 2 y42 ,
C = 1 + 4||
(4.43)
and are to be understood as functions of u through the identifications (4.39) and (4.40). In
Brinkmann coordinates, the metric is given by

ds 2 = 2 du dv + d r32 + dx 2 + d y42 + Fr r32 + Fx x 2 + Fy y42 du2 ,
(4.44)
with

a2 1
,
Fr = J 1 +
(1 a cos 2J u)2

a2 1
2
,
Fx = J 1 3
(1 a cos 2J u)2

2 J 2.
Fy = 1 + 4||
2
(4.45)
Note that, since 0 < a < 1 the metric is no-where singular. The remaining nonzero fields are
e2 = gs2 ,
4J
F3 =
gs
J
F5 = du (dr1 dr2 dr3 dx dy1 dy2 dy3 dy4 ).
gs
(4.46)
These are the same as in Eq. (4.18) for the same geodesic at zero rotation. This is in agreement
with the results of [27], corresponding to the limiting case = = 0, where it was noted that
the RR 5-form field strength at the Penrose limit for the (J, 0, 0) geodesic retains the same form
77
as in the zero-rotation case. Indeed, since the Penrose limit and the -deformation procedure
commute, the fact that in the undeformed case the Penrose limits of the metric along the torus
directions, of the dilaton and of the RR 5-form are the same as at zero rotation guarantees that,
in the deformed case, the fields in (4.46) will be equal to those in (4.18). The deformation affects
only the function Fy in the metric and the NSNS and RR 3-form field strengths while the effect
of nonzero rotation parameters manifests itself only in the functions Fr and Fx in the metric.
(0, 0, J ) geodesic. The differential equation for r becomes
r 2 = 1 J 2 r 2 2 ,
(4.47)
with as in (4.38). Its solution is

b2
r 2 (u) = 2 sin2 J u, b 1 + J 2 r02 ,
J
from which it follows that

b2 1
(u) = 1
.
b2 sin2 J u
(4.48)
(4.49)
The Penrose limit of the deformed metric reads

ds 2 = 2 du dv + Ar d r32 + Ax dx 2 + Ay d y22 + A y d y 22 C du2 ,
(4.50)
where
d y22 = dy12 + dy22 = d12 + 12 d12 ,
d y 22 = dy32 + dy42 = d22 + 22 d22 ,
(4.51)
and
1
J 2 r 2 ,
Ay = ,

2

y2
y
2 y22 + 2
+ y 22 + 4||
C = J2
,

Ar = r 2 ,
Ax =
1
A y =
,

and again are to be understood as functions of u through (4.47) and (4.48). In Brinkmann coordinates, the metric reads

ds 2 = 2 du dv + d r32 + dx 2 + d y22 + d y 22 + Fr r32 + Fx x 2 + Fy y22 + Fy y 22 du2 , (4.52)
with

2
b2 1
b 1
2
+
3
,
b
4(1 b2 cos2 J u)2 sin2 J u

1
b2 1
b2 1
5
,
Fx = J 2 1
4 (1 b2 cos2 J u)2 4 sin2 J u(1 b2 cos2 J u)

(b2 1)(4b2 cos4 J u 2 (b2 + 1) cos2 J u)
b2 sin2 J u
2
2
Fy = J
+ 4|| ,
1 b2 cos2 J u

1 b2 cos2 J u (b2 1)(4b2 cos4 J u 2 + (1 3b2 ) cos2 J u)
2 ,
+
+
4|
|
Fy = J 2
b2 sin2 J u
(4.53)

Fr = J 2 1 +
78
while the remaining nonzero fields are

e2 = gs2 ,
4J
F3 =
gs
1 b2 cos 2J u
J
du
F5 =
2gs b sin J u 1 b2 cos2 J u

(dr1 dr2 dr3 dx dy1 dy2 dy3 dy4 ).
(4.54)
In contrast to the previous case, these differ from the corresponding expression for the same
geodesic at zero rotation in that the RR 5-form now has an r0 -dependent overall coefficient.
Again this is in agreement with the corresponding analysis of [27] for the limiting case =
= 0. Now, the deformation affects only the functions Fy and Fy in the metric and the NSNS
and RR 3-form field strengths while the effect of turning on rotation parameters manifests itself
in all F -functions in the metric and in the RR 5-form field strength.
(0, J, J ) geodesic. Now, it is convenient to introduce
G 1 =
2 )r 2 r 2
(1 + ||
0
r 2 r02
H=
(1 + 2 )r 2 r02
r 2 r02
(4.55)
which are the effective values of the functions in (3.27) for the given ansatz, now being functions
of r. The differential equation for r has the form
J 2 r 2 2
,
(4.56)
G
with as in (4.38). Unfortunately, this equation cannot be solved explicitly in the general case,
for which we will content ourselves with presenting the Penrose limit only in the Rosen-like
form, with the u-dependence of the metric components entering implicitly through (4.54). This
metric is given by
Hr 2 = 1
ds 2 = 2 du dv + Ar d r32 + Ax dx 2 + Ay d y22 + By (y1 dy2 y2 dy1 ) du

+ A d 2 + A d 2 + B d du C du2 ,
(4.57)
where
d y22 = d12 + 12 d12 ,
(4.58)
and

Ar = H1/2 r 2 ,
Ay = H1/2 ,
Ax = H1/2

G
J 2 r 2 ,

2 J,
B2 = 4||
GH1/2
A w =
,
B = 4J,

2 4||
4 )
22]
y22 4||
J 2 [(1 ||
.
C=
H1/2
A =
H1/2
,

(4.59)
79
On the other hand, for the case of a pure -deformation, Eq. (4.56) can be solved exactly with
the result

b2 G0
Ju
r 2 (u) = 2 sin2 1/2 ,
(4.60)
J
G0
where b is as in (4.48) and
G01 = 1 + 2 .
(4.61)
The various functions for the metric in Rosen-like coordinates are found by substituting (4.60)
into (4.57) and setting G G0 and H 1. In Brinkmann-like coordinates, the formulas for
the various functions are rather messy and we refrain from quoting them. We just note that the
solution has F1 = F3 = 0 and that all F -functions in the metric and all nonzero fields are affected
both by the deformation and by the rotation parameters.
4.3.3. One rotation parameter (disc)
We finally consider the case of one rotation parameter. As remarked earlier on, when = 0,
the solutions for one rotation parameter are obtained from those for two equal rotation parameters through the replacement r02 r02 . Making this replacement in the corresponding Penrose
limits, we find the following results:
(J, 0, 0) geodesic. The Penrose limit of the solution in Brinkmann coordinates is given by
Eqs. (4.44)(4.46), now with

a = 1 + 4J 2 r02 .
(4.62)
Note that now a is real for all values of J r0 . Also, since a > 1 the PP-wave metric is singular at
cos 2J u = 1/a.
(0, 0, J ) geodesic. The Penrose limit of the solution in Brinkmann coordinates is given by
Eqs. (4.52)(4.54), now with

b = 1 J 2 r02 .
(4.63)
(0, J, J ) geodesic. In the general case, the Penrose limit of the deformed metric in Rosenlike coordinates is given by Eq. (4.57) with d y22 given in (4.58) and the A-, B- and C-functions

given in (4.59) but with replaced by + =
1+
r02
r2
and with G 1 and H appropriately mod-
ified. For the case of a pure -deformation, there exists the explicit solution (4.58) for r(u) and
the Penrose limit in Rosen-like coordinates is found by substituting that solution into Eqs. (4.57)
(4.59), with replaced by + .
5. Giant gravitons on -deformed PP-waves
Given the marginally-deformed geometries, it is interesting to investigate the various extended
objects that they can support. In this respect, an important role is played by BPS configurations
of spherical D3-branes, the so-called giant gravitons. To summarize the basic facts, the authors
of [29], building on results of [30], considered a KK excitation (graviton) in AdS5 S5 with
nonzero angular momentum along an S5 direction and contemplated the possibility that it might
80
blow up on an S3 inside the S5 without raising its energy. They found that such a state (the giant
graviton) can indeed exist, with the blowing up being due to its angular momentum and the extra
force required to keep it stable under shrinking being provided by RR repulsion. Soon after that,
it was found [31] that there also exist dual giant gravitons with similar properties, supported
on the S3 inside the AdS5 part of the geometry. The construction of giant gravitons has been
extended towards various directions in [32], while investigations from the dual field-theory side
have been carried out in [33]. An interesting result is that, in certain cases [34,35] where the
geometry of the S5 supporting the giant graviton is deformed, the latter has higher energy than
the point graviton, and may even not exist at all as a solution.
This latter fact serves as a motivation for examining giant gravitons in the marginallydeformed solutions of interest, since one of the main features of the latter is precisely a deformation of S5 . In the existing literature, giant gravitons have been considered only for the case
of -deformations. For the full deformed solutions, giant gravitons were first constructed in [35]
for a special case of the non-supersymmetric three-parameter background of [4,5] and, more
recently, for the general three-parameter [36] and the single parameter [36,37] backgrounds;
in particular, the giant gravitons of [36,37] were found to be independent of the deformation
parameter and hence still degenerate with the point graviton. For the PP-wave limits of the
marginally-deformed backgrounds, giant gravitons have been constructed, in analogy to the considerations of [38] for the PP-wave limits of AdS5 S5 , in [39] for the two PP-wave limits of
-deformed AdS5 S5 , namely those along the (J, 0, 0) and the (J, J, J ) geodesic. For the first
geodesic, the solution was constructed exactly and it was found that -deformations do not lift
the degeneracy of the giant and point gravitons, in accordance with [37], and a stability analysis
indicated that the former is perturbatively stable. For the second geodesic, the problem was attacked perturbatively in , with the leading-order analysis indicating that the -deformation lifts
the degeneracy in favor of the point graviton, but no definitive conclusion on whether the giant
graviton survives the deformation for large enough was reached.
Here, we address the question of identifying giant gravitons on the PP-wave limit of the
deformed geometries, this time in the presence of -deformations and/or nonzero rotation parameters. Restricting to PP-waves along the (J, 0, 0) geodesic, our exact analysis for the giant
graviton residing on the deformed S5 part of the geometry shows that -deformations have an
altogether different effect than that of -deformations, lifting the degeneracy of the giant and
point gravitons and, for above a critical value, completely removing the giant graviton from
the spectrum. Moreover, the analysis of small fluctuations of the giant graviton reveals that the
latter is perturbatively stable throughout its range of existence. We also consider dual giant gravitons residing on the AdS5 part of the geometry, in which case the deformation does not affect
neither the solution, nor its stability properties. In what follows, we present the relevant analysis,
keeping a similar notation with [39] in order to facilitate comparison.
5.1. Giant gravitons on the deformed PP-waves
To describe the giant-graviton solutions in the PP-wave spacetimes of interest, we consider
the action for a probe D3-brane in this background, given by the sum of the DiracBornInfeld
and WessZumino terms,
SD3 = SDBI + SWZ

4
= T3 d e
det P[G B] + T3
P Aq eB2 ,
q
(5.1)
81
where T3 is the D3-brane tension, equal to 1/(2)3 in units where = 1, and P[f ] stands for
the pullback of a spacetime field f on the worldvolume. Below we describe the construction of
giant gravitons and dual giant gravitons on the deformed PP-waves under consideration.
5.1.1. Giant gravitons
Starting with ordinary giant gravitons, we want to describe a D3-brane wrapping the S3 inside
the S5 in the deformed geometry. To do so, we employ the gauge choice
= u,
1 = ,
2 = 2 ,
3 = 3 ,
(5.2)
and we consider the ansatz

v = u,
r4 = 0,
= 0 = const.
(5.3)
Noting that the spatial brane coordinates are just the angular coordinates employed in the Rosen
form (4.12)(4.15) of the PP-wave solution of interest, the pullbacks of the various fields on the
D3-brane can be immediately read off from these equations. Setting for convenience J = 1, we
find

2 02 , 02 , 02 sin2 , 02 cos2 ,
P[G] = diag 2 1 + 4||
(5.4)
and

P[B2 ] = 2 02 du sin2 d2 cos2 d3 ,

2
P[A2 ] = 02 du sin2 d2 cos2 d3 ,
gs
1
P[A4 ] = 04 cos sin du d1 d2 d3 .
gs
From these equations, we readily compute

det P[G B] = 2 + 1 + 4 2 02 06 cos2 sin2 .
(5.5)
(5.6)
As in [39], we note that all dependence has dropped out due to the cancellation of terms
coming from the metric and from B2 . The important fact is that, since the metric now involves
2 in place of 2 while B2 still depends only on , there remains a non-trivial dependence on
||
the deformation through the parameter . Noting also that, since P[B2 A2 ] = 0, only P[A4 ]
contributes to the WessZumino action, we write the full D3-brane action as

T3
du d3 03 2 + 1 + 4 2 02 04 = du LD3 ,
SD3 =
(5.7)
gs
where d3 = cos sin d d1 d2 is the S3 volume element and LD3 is the Lagrangian

LD3 = M 03 2 + 1 + 4 2 02 04 ,
(5.8)
where
M=
2 2 T3
1
N
=
=
.
gs
4gs R 4
(5.9)
82
Since LD3 is independent of v and u, we have two conserved first integrals, given by the lightcone momentum
P =
M03
LD3
,
=
2 + (1 + 4 2 )02
(5.10)
and the light-cone Hamiltonian

E=
M 3 [ + (1 + 4 2 )02 ]
LD3
M04 .
LD3 = 0
2
2
2 + (1 + 4 )0
(5.11)
Solving (5.10) for and substituting in (5.11), we write

(1 + 4 2 )P 2
M2 6
0 M04 +
0 .
2P
2
As a function of 0 , E has local extrema at the radii

(2 )P
0 = 0 =
,
0 = 0,
3M
where we have defined

1 12 2 .
E=
(5.12)
(5.13)
(5.14)
The corresponding light-cone energies are given by

P2
.
27M
Note also that (5.10) and the second of (5.13) lead to the constraint
E0 = 0,
E = (2 )2 (1 )
(5.15)
2 2
= 0
(5.16)
(2 )(1 ),
9
determining in terms of 0 .
1
, the radius 0 = 0 corresponds to a local maximum, while the radii 0 = 0
For 0 <
2 3
and 0 = 0+ correspond to two local minima, the point graviton and the giant graviton. At exactly = 0, we recover the usual result that E+ = E0 , i.e., that the giant graviton is degenerate
1
in energy with the point graviton. However, for 0 < <
, we have E+ > E0 , i.e., the degen2 3
1
eracy is lifted with the giant graviton becoming energetically unfavorable. At =
, the radii
2 3
0 = 0 and 0 = 0+ degenerate into a saddle point, leaving only one minimum at 0 = 0.
1
, the only extremum is the minimum at 0 = 0: the giant graviton disappears
Finally, for >
2 3
from the spectrum. The situation is depicted in Fig. 1. To summarize, for the PP-wave along the
(J, 0, 0) geodesic, complex -deformations, unlike -deformations, have the effect of lifting the
degeneracy of the giant and point gravitons for small values of and of removing the giant graviton from the spectrum for large values of . Also, as we shall see explicitly later on, although
the effective Lagrangian (5.8) depends only on , the spectrum of small perturbations about the
giant graviton solution is dependent on both deformation parameters.
We note that the above results remain valid when the rotation parameter r0 is turned on. This
follows from the fact that the relevant components of the metric and the remaining fields are
identical to those at zero rotation (see the comments following Eq. (4.46)).
83
Fig. 1. Light-cone Hamiltonian for the PP-wave along the (J, 0, 0) geodesic plotted as a function of . The three curves
(dashed) and = 1.1
(dotted). The plots are shown in units where M = P = 1.
correspond to = 0 (solid), = 0.5
2 3
2 3
5.1.2. Dual giant gravitons

Proceeding to the case of dual giant gravitons, we have to consider a D3-brane wrapping
the S3 originating from the AdS5 part of the geometry. Since the latter part of the geometry is
unaffected by the deformation, it is immediately seen that the dual giant graviton solution exists
and is independent of the deformation; however, to set up the notation for the stability analysis
that follows, let us demonstrate it explicitly. To do so, we first parametrize, in analogy to (4.13),
2 , 3 ) defined by
the coordinate vector r4 by the coordinates (,
,

d r42 = d 2 + 2 d 2 + sin2 d 22 + cos2 d 32 ,
(5.17)
we employ the gauge choice
= u,
1 = ,
2 = 2 ,
3 = 3 ,
(5.18)
and we consider the ansatz

v = u,
y4 = 0,
= 0 = const.
(5.19)
Proceeding as before, we find that the only relevant pullbacks of the spacetime fields are

02 cos2 ,
P[G] = diag 2 02 , 02 , 02 sin2 ,
(5.20)
and
P[A4 ] =
1 4
cos sin du d1 d2 d3 ,
gs 0
and we arrive at the action

SD3 = du LD3 , LD3 = M 03 2 + 02 04 ,
(5.21)
(5.22)
with M as in (5.9). The dual giant graviton solution is found as before and is obviously independent of the deformation, which implies that it is degenerate with the point graviton for all values
of the deformation parameters. However, as we shall see, this degeneracy does not extend to the
spectrum of fluctuations around this solution.
84
When the rotation parameter r0 is turned on, however, the above solution is no longer valid.
This is because the SO(4) symmetry of the r4 = (r3 , x) directions is broken by the rotation
parameter, as manifested by the fact that the functions Fr (u) and Fx (u) in (4.44) are different.
5.2. Small fluctuations and perturbative stability
We now turn to an analysis of small (bosonic) fluctuations about the giant graviton configurations in deformed PP-waves, following the treatment of [40]. We consider both ordinary and
dual giant gravitons and, for the former case, we also take account of the presence of rotation. As
we shall see, although the spectrum of fluctuations is affected by the deformation, the standard
result that these configurations are perturbatively stable remains unchanged.
5.2.1. Giant gravitons
Starting with ordinary giant gravitons, the perturbation of the classical configuration is described by keeping the gauge choice as in (5.2) and perturbing the embedding as
v = u + v(u, , 2 , 3 ),
r3 = r3 (u, , 2 , 3 ),
= 0 + (u, , 2 , 3 ),
x = x(u, , 2 , 3 ).
(5.23)
Computing the pullbacks of the various fields as before, inserting them in the D3-brane action
and expanding up to second order in the fluctuations, we write
SD3 = S0 + S1 + S2 + ,
(5.24)
where the various terms correspond to the respective powers of the fluctuations. The zeroth order
is just the classical action given by (5.7) and (5.8). The linear term reads
S1 =

M02
2 2 2 + (1 + 4 2 )02

du d3 0 u v 2 3 + 2 1 + 4 2 02 20 2 + 1 + 4 2 02 , (5.25)
and it depends only on the deformation parameter . The first term is a total derivative that
vanishes upon integration over u, while the second term vanishes upon imposing the constraint
(5.16). Finally, calculating the quadratic term and making use of (5.16) (with the upper signs),
we obtain

3M02
8(1 ) 2
du d3
Fr (u)r32 Fx (u)x 2
S2 = 2
3
4 (2 + )

2
12
9
u v + 4 2 (2 3 )r5
(u v)2 (u r5 )2
(2 + )0
(2 + )2 02

1
(2 + )2
+ 2 h v v +
(5.26)
h r5 r5 .
9
0
Here, is the -dependent quantity defined in (5.14), h is the metric on S3 and we introduced
the shorthand r5 = ( r3 , x, ). Also, Fr (u) and Fx (u) are the u-dependent functions defined
in (4.45), with the parameter a given by the second of (4.39) and by (4.62) for the case of two
and one rotation parameters respectively. In the limit of zero rotation both Fr and Fx equal to
85
1. We note that, unlike the zeroth order and linear terms, the quadratic term has some, albeit
restricted, dependence on the deformation parameter in addition to the dependence on . After
several integrations by parts, the quadratic action becomes

3M02
8(1 ) 2
du d3
S2 = 2
Fr (u)r32 Fx (u)x 2
3
4 (2 + )

2

12
1
3
2
u v + 2 v
u S3 v
(2 + )0
2+
0

2+ 2
2
2
2
+ r5 u
(5.27)
S3 4 (2 3 ) r5 ,
3
where S3 is the Laplacian on S3 . To proceed, we may expand all fluctuations in the basis
spanned by the combinations S3 ,n2 n3 (, 2 , 3 ) of S3 harmonics having definite quantum
numbers under the two U (1)s corresponding to shifts of 2 and 3 , i.e., by the simultaneous
eigenfunctions of the operators S3 and 2,3 with
S3 S3 ,n2 n3 (, 2 , 3 ) = ( + 2)S3 ,n2 n3 (, 2 , 3 ),
2,3 S3 ,n2 n3 (, 2 , 3 ) = in2,3 S3 ,n2 n3 (, 2 , 3 ),
(5.28)
where , n2 and n3 are required to satisfy4

= 0, 1, . . . ,
|n2 | |n3 | = 2k,
k = 0, 1, . . . .
(5.29)
For the v and fluctuations, the fact that their coefficients in the action are u-independent
allows us to introduce an eiu time dependence and thus we can write
v(u, , 2 , 3 ) = vn2 n3 eiu S3 ,n2 n3 (, 2 , 3 ),
(u, , 2 , 3 ) = n2 n3 eiu S3 ,n2 n3 (, 2 , 3 ),
(5.30)
while for the (r3 , x) fluctuations, we can only set

r3 (u, , 2 , 3 ) = r3,n2 n3 (u)S3 ,n2 n3 (, 2 , 3 ),
x(u, , 2 , 3 ) = xn2 n3 (u)S3 ,n2 n3 (, 2 , 3 ).
(5.31)
The spectrum of fluctuations follows from inserting the expansions (5.30) and (5.31) in the
equations of motion stemming from the action (5.27). Starting from the null fluctuation v and
the radial fluctuation , we find that they satisfy the coupled system
(2+)2 (+2)92
6i

(2+)
22
0
(2+)
0
vn2 n3 = 0, (5.32)
n2 n3
(2+)2 (+2)92 +36 2 (n2 n3 )2 +24(1)
6i
(2+)0
which leads to the following spectrum

2+ 2
(2 + )
2
2
2
,n2 n3 =
( + 2) + 2
+ (n2 n3 )
3
3
4 The explicit expression for the
S3 ,n2 n3 s can be written in terms of Jacobi polynomials (see, for instance, Section 6
of the first of [14]).
86

2 2
2+
3
2
( + 2) +
(2 + )
+ 2 (n2 n3 )2
3
2
.
(5.33)
2
2
are positive-definite while, for = 0 ( = 1), the ,n
are positiveObviously the +,n
2 n3
2 n3
semidefinite, with a zero mode occurring for = 0. Therefore, the only potential source of
2
instabilities for these fluctuations is one of the ,n
becoming negative for some value of .
2 n3
This would only be possible if the inequality

P (x) ( + 2) 24 x 2 + 4 ( + 2) + 6 x + 4 ( + 2) + 9 2 (n2 n3 )2 < 0,
(5.34)
could be satisfied for some x with 0 x 1. It is easily seen that this cannot happen for any values of . Turning to the (r3 , x) fluctuations, we find that they satisfy the Schrdinger equations

2+ 2
d2
( + 2) 4 2 (n2 n3 )2 r3,n2 n3 (u) = 0,
2 + Fr (u)
(5.35)
3
du
and

2+ 2
d2
2
2
+
F
(u)
(
+
2)
(n
n
)
xn2 n3 (u) = 0.
x
2
3
3
du2
(5.36)
To examine them, we may consider the following cases.

Zero rotation (r0 = 0). In the absence of rotation parameters (in which case Fr = Fx =
1), the r3 and x fluctuations obey the same Schrdinger equation with a u-independent
potential. Introducing an eiu dependence, we easily obtain the spectrum

2+ 2
2
2
2
=
1
+
4
(n
n
)
+
( + 2),
r,n
(5.37)
2
3
2 n3
3
2
are manifestly positive-definite, signifying stability
whence we verify that the r,n
2 n3
2
against small perturbations in these directions. The r,n
are increasing functions of
2 n3
but decreasing functions of with the last term ranging from ( + 2) at = 0 to 49 ( + 2)
1
. Note also that no zero mode is possible.
at =
2 3
Nonzero rotation parameters (r0 = 0). In this case, the Schrdinger equations (5.35)
and (5.36) have periodic potentials Fr and Fx , respectively, given by (4.45) (with the parameter a in (4.39)) and fixed eigenvalue depending on the deformation parameters and
as well as on the quantum numbers and n2,3 .5 From the shape of the potentials and
since the fixed eigenvalue is non-negative, we infer stability. However, since the potentials
are periodic the spectrum is continuous with mass gaps, which tend to zero as the parameter
a 0. Hence, for fixed quantum numbers and n2,3 , there exist ranges of the deformation
parameters and for which any other solution than the vanishing one is not allowed.
In conclusion, despite the fact that -deformations render the giant graviton states energetically
unfavorable, the small-fluctuation analysis indicates that these objects are perturbatively stable
5 Exactly the same equations appeared in [27] in the light-cone quantization of strings moving in these PP-wave
backgrounds (for = = 0).
87
in the range of where they are allowed to exist in the first place. Presumably, the effect of deformations renders the giant gravitons metastable rather than unstable. To further investigate
this aspect, one may seek bounce-like instanton solutions connecting the giant and the point
graviton and calculate the tunneling probability by standard WKB methods. In doing so, one
must appropriately take into account possible fermionic zero modes resulting from the breaking
of supersymmetries by the instanton, which tend to suppress the tunneling rate.
5.2.2. Dual giant gravitons
We next consider dual giant gravitons which, as we recall, are independent of the deformation.
Now, the perturbation can be described by keeping the gauge choice as in (5.17) and perturbing
the embedding according to
2 , 3 ),
v = u + v(u, ,
2 , 3 ).
y4 = y4 (u, ,
2 , 3 ),
= 0 + (u,
,
(5.38)
Repeating the same steps as before, we find that the quadratic term in the action of the fluctuations
reads

M 02
4
2 y42
S2 =
du d 3 1 + 4||
u v
0
4 2

2
1 2
+ 2 v u S3 v + y5 u S3 y5 ,
(5.39)
0
where we introduced the shorthand y5 = ( y4 , ).
We see that the deformation enters only
through a modification of the mass term of y4 . Introducing an eiu time dependence and
expanding the fluctuations on S3 as before, we find the fluctuation spectrum

2
,n
(5.40)
= ( + 2) + 2 2 ( + 2) + 1,
2 n3
and
2
2 + ( + 2),
= 1 + 4||
y,n
2 n3
(5.41)
for the (v, )

and y4 fluctuations respectively. This spectrum is manifestly positive2
semidefinite, with ,n
having the expected zero mode for = 0. We conclude that the
2 n3
deformation does not affect the stability of dual giant gravitons, its sole effect being just a raise
of the energy of the y4 fluctuations.
6. Probing the deformed geometry with Wilson loops
To further investigate the effects of -deformations we now turn to another, completely different, direction, of more phenomenological nature. Namely, we consider the potential for a static
heavy q q pair in the dual gauge theory, extracted from the expectation value of a rectangular
Wilson loop extending along the Euclidean time direction and one space direction. On the gravity side, the Wilson loop expectation value is calculated [15,41] by minimizing the NambuGoto
action for a fundamental string propagating into the dual supergravity background, whose endpoints are constrained to lie on the two sides of the Wilson loop. Below, we first briefly review the
procedure for calculating Wilson loops in general supergravity backgrounds of interest [14,42]
and then we apply it to the case of -deformations of the Coulomb branch.
88
6.1. General formalism

As stated above, the calculation of a Wilson loop in the gravity approach amounts to extremizing the NambuGoto action (taking into account the contribution of the NSNS 2-form
if necessary) for a string propagating in the dual geometry whose endpoints trace the loop.
To describe the propagation of the string, we first fix reparametrization invariance by taking
(, ) = (t, x). We next need to find a suitable ansatz that is sensitive to -deformations. Specializing to a radial trajectory, it is easily seen that the only available choice for the embedding
is6
r = u(x),
= 0,
= ,
4
2 = 3 = const,
rest = const.
1 = const,
(6.1)
We next pass to the Euclidean using the analytic continuation t it .7 Then, the NambuGoto
action is found to be

T
dx f (u)/R 4 + g(u)u 2 ,
S=
(6.2)
2
where the prime denotes a derivative with respect to x and
f (u) = R 4 Gtt Gxx ,
g(u) = Gtt Guu .
(6.3)
We do not need to consider at all the contribution of the NSNS 2-form, since it is vanishing due
to the ansatz (6.1) and to the fact that, for the extremal D3-brane distributions considered here,
the 2-forms C3i generating terms for the NSNS 2-form through (2.8) vanish identically.
Since the action (6.2) does not explicitly depend on x, it leads to the first integral u0 , identified
with the turning point of the solution. Solving the corresponding first-order equation for x in
terms of u we find that the linear separation of the quark and antiquark is

g(u)
.
L = 2R 2 f 1/2 (u0 ) du
(6.4)
f (u)[f (u) f (u0 )]
u0
The energy of the configuration is given by the action (6.2) divided by T . Subtracting the selfenergy contribution, we obtain

u0

g(u)f (u)
1
1
du
du g(u),
g(u)
E=
(6.5)
f (u) f (u0 )
u0
umin
where umin is the minimum value of u allowed by the geometry. In specific examples, we are
supposed to solve for the auxiliary parameter u0 in terms of the separation distance L. Since this
cannot be done explicitly except for some special cases, in practice one regards Eq. (6.4) as a
parametric equation for L in terms of the integration constant u0 . Combining it with Eq. (6.5)
for E, one can then determine the behavior of the potential energy of the configuration in terms
of the quarkantiquark separation.
6 If, for instance, = /2, then the resulting Wilson loop potentials are identical to those for the N = 4 undeformed
theory computed in [42]. Also, to conform with standard notation in the literature, we use u instead of r in the Wilsonloop computations.
7 In extending this to the finite-temperature case, one also has to take r ir due to the presence of nonzero metric
0
0
components Gti r0 dt di .
89
6.2. Application: -deformations of the Coulomb branch

As an application of the above, we extend the results of [42] for the undeformed theory, by
considering the behavior of the static q q potential for the case of pure -deformations ( = 0) of
the Coulomb branch of the gauge theory at zero temperature. This was previously examined in
[16], where most of the qualitative features of the potentials were extracted numerically. Here, we
pursue a more careful analysis, which allows us to discover certain features previously unnoticed.
For the analysis that follows, it is convenient to use the single dimensionful parameter r0 of
the theory to switch to dimensionless variables. To do so, we set
u r0 u,
u0 r0 u0 ,
(6.6)
and
R2
r0
(6.7)
L,
E E.
r0
Also, to keep the discussion at a reasonable length, we restrict to the cases of two equal rotation
parameters and one rotation parameter. The results are presented below.
L
6.2.1. Two equal rotation parameters (sphere)

For the case of two equal rotation parameters, the effective metric for the trajectories (6.1)
reads

ds 2 = H1/2 H 1/2 dt 2 + dx 2 + H1/2 H 1/2 du2 ,
(6.8)
with
H=1+
2 u2
,
u2 r02
H=
R4
.
u2 (u2 r02 )
(6.9)
Performing the analytic continuation, calculating the functions f and g according to (6.3), inserting into (6.4) and (6.5) and switching to dimensionless variables, we find the following exact
expressions for L and E in terms of complete elliptic integrals8

du
1
2

L = 2u0 u0
1 + 2
2
u (u2 u )(u2 1)(u2 + u2 1 )
0
u0
=
and
2u0
(u20
1
)(2u20
1+ 2
1
)
1+ 2

2
a , k K(k) ,
1+ 2

!

u2 1+1 2 !
u2 (u2 1+1 2 )
"
2
E = 1 +
du
1
u2 1
(u2 u20 )(u2 + u20 1+1 2 )
u0

u0

u2 1+1 2
2
1 +
du
u2 1
1
8 In the following we adopt the notation and use properties of elliptic integrals as in [43] and [44].
(6.10)
90

1 + 2
2u20

1 2
2
a
K(k)
E(k)
+
E(c)
K(c)
,
1 + 2
1 + 2
a2 =
(6.11)
where
k =
2
2
1+ 2
2u20 1+1 2
u20 +
u20
1
1+ 2
2u20 1+1 2
1
c=
.
1 + 2
(6.12)
The resulting plots of E versus L are shown in Fig. 2. For = 0, the behavior is that for the
undeformed case [42]. As is turned on, the length and energy curves closely resemble the
van der Waals isotherms for a statistical system with u0 , L and E corresponding to volume,
pressure and Gibbs potential respectively (see, for instance, [45]). In the region below a critical
point, < cr , the behavior is analogous to that of the statistical system at T < Tcr . Namely,
the potential energy E (i) starts out Coulombic at small distances, (ii) becomes a triple-valued
function of L with the state of lowest energy corresponding to the initial branch, (iii) passes a selfintersection point after which it becomes a triple-valued function of L with the state of lowest
energy corresponding to the second branch and (iv) returns to being a single-valued function
of L with approximately linear behavior. By standard arguments, the physical path in the length
and energy curves must correspond to the physical isotherms of the statistical system, with the
self-intersection point in the energy curve indicating a first-order phase transition with order
parameter u0 . In the region above the critical point, > cr , the behavior is analogous to that
of the statistical system at T > Tcr . Now the energy is single-valued throughout and the firstorder phase transition has degenerated into a second-order one between a Coulombic phase and
(a)
(b)
(c)
Fig. 2. Energy as a function of length for the case of two equal rotation parameters.
91
a confining phase with a linear potential, as we will show below. The above critical behavior is
similar to that found in [42] for a different system, namely for the undeformed theory at finite
temperature and nonzero chemical potential. Since the two cases are in complete analogy, we
refer the reader to that work for several related computational details.
To determine the critical value of , we consider the derivative of the separation length L with
respect to u0 . This derivative is proportional to the function
f (u0 ; ) =
2

1
+ 2 1 + 2 u20
2
1 + 2
2

1 + 2 u20 2u20 K(k) + 1 u20 E(k) .
(6.13)
This function has a single zero corresponding to a global maximum of the length for = 0, two
zeros corresponding to a local minimum and a local maximum of the length for 0 < < cr ,
and no extrema for > cr . To calculate cr , we proceed by expanding f (u0 ; ) around u0 = 1
corresponding to the modulus k = 1; although this procedure is not a priori valid, it will be
justified by our final result. We have

f = 2 4x 8 + 15 2 + 2 + 2 ln x + O x 2 ,
x
1 + 2
(u0 1),
8(1 + 2 2 )
(6.14)
and

f
= 4 10 + 16 2 + 2 + 2 ln x + O(x).
x
Setting f = 0 gives the transcendental equation
a ln a =
8+15 2
2
2+ 2 e 1 ,
e
4(2 + 2 )
a = xe
8+15 2
2+ 2
(6.15)
(6.16)
This equation has two solutions a1 , a2 if < cr , one solution a1 = a2 = ac if = cr and no

solutions if > cr . The critical value c is obtained when the above inequality is saturated
in which case the solution is a = e1 . It turns out that in this case we have in addition that
f/x = 0. The corresponding transcendental equation is
zez+5 = 44,
z 11
cr2
.
2 + cr2
(6.17)
This gives z 0.235, whence cr 0.209, which clearly is small enough to validate in retrospect
the approximation method we used to compute it. Having a solution to (6.16) we obtain from the
definition in (6.14) that the corresponding critical value(s) for u0 are given by
u0i = 1 +
ai 1 + 2 2 8+1522
e 2+ ,
8 1 + 2
i = 1, 2,
(6.18)
for cr .
Finally, note that for u0 1 (large L) we recover the usual Coulombic behavior, while for
u0 1 we find the asymptotics
1
ln
,
L
2
u
01
1 + 2
1
2
ln
E
.
2
u
01
2 1 + 2
(6.19)
92
Combining these expressions, we obtain the potential
L,
(6.20)
2
which demonstrates the linear confining behavior claimed earlier on, as long as > 0. This linear
potential was also found in the studies of [16] where, however, the critical behavior found here
was missed. Using the first of (6.19) and reinstating dimensional units according to the first (6.7),
2
we find that confinement sets in at length scales L 2 Rr0 .
E
1+2
6.2.2. One rotation parameter (disc)

For the case of one rotation parameter, the effective metric for the trajectories (6.1) reads

ds 2 = H1/2 H 1/2 dt 2 + dx 2 + H1/2 H 1/2 du2 ,
(6.21)
with
H=1+
2 u2
,
u2 + r02
H=
R4
.
u2 (u2 + r02 )
(6.22)
Proceeding in the same way as before, we find that the length and potential energy are given by

L = 2u0
=
u20
1
+
1 + 2
E=
du
u (u2 u20 )(u2 + 1)(u2 + u20 +
u0
2u0
(u20
1
)(2u20
1+ 2
and

1 + 2
u2 +
1
)
1+ 2
1
1+ 2

2
a , k K(k) ,
!
!
"
1
)
1+ 2
u2 (u2 +
1
)
1+ 2
2
2
2
2
(u u0 )(u + u0 + 1+1 2 )
(6.23)
1
u2 + 1
u0

u0

u2 + 1+1 2
2
du
1 +
u2 + 1
0

1 2
1
2
2
= 1 +
a K(k) E(k) + E(c)
2u0 +
K(c) ,
1 + 2
1 + 2
du
(6.24)
where
k =
2
2
1+ 2
2u20 + 1+1 2
u20
a =
2
u20 +
1
1+ 2
2u20 + 1+1 2
c=
.
1 + 2
(6.25)
Fig. 3 shows the resulting plots of E versus L. For

u0 1 (small L) we recover the standard
Coulombic behavior [15] enhanced by the factor 1 + 2 , while for u0 0 we find the asymptotics
L 2E(i )u0 ,
1
E E(i )u20 ,
2
(6.26)
93
Fig. 3. Energy as a function of length for the case of one rotation parameter, shown for = 0 (solid) and > 0 (dashed).
which reduce to those in [42] for = 0. Combining these expressions, we obtain

E
( L)2
,
8E(i )
(6.27)
which shows that there is complete screening at the screening length Lc = that is invariant
under -deformations and the same as the one found [42] for the undeformed case. The effect of
-deformations is to enhance the quarkantiquark force for small separations and to suppress it
for separations close to the screening length.9 All of our results reduce smoothly to those in [42]
for = 0.
We finally remark that it would be very interesting to examine the stability of the string trajectories used for calculating the quarkantiquark potential in the deformed theories. For instance,
in the undeformed theory with two rotation parameters, the potential for our trajectory (6.1) with
= 0 (shown in Fig. 2(a)) is a double-valued function while the potential for a trajectory with
= /2 (which is insensitive to deformations and thus has not been considered here) exhibits a
confining behavior at large distances [42]. It was found in [46] that the upper branch in the = 0
case as well as the region giving rise to linear behavior in the = /2 case are actually unstable
under small fluctuations, the latter fact being in accordance with our physical expectation about
the absence of confinement in N = 4 SYM. In the presence of deformation, the potential for
= 0 gives the behavior shown in Fig. 2(b), (c), while the potential for = /2 stays invariant.
It is then important to ask whether the confining regions of these potentials are stable, as the
N = 1 supersymmetry of the -deformed theories actually leads us to expect a confining behavior at large distances, at least in some regions of the moduli space. We note that the question of
stability is meaningful even in the = /2 case since, although the classical string solution is
independent of the deformation, small fluctuations about it are not. We hope to report on work in
that direction in the future [47].
9 Note the crossover behavior for the curves with = 0 and > 0 in our Fig. 3 at a certain length. This is understood
by our analytic result (6.26) as well as the enhanced Coulombic behavior in the UV noted above. In that respect we
disagree with the shape of the potential presented in Fig. 3 of [16].
94
7. Conclusions
In this paper, we have explicitly applied the LuninMaldacena construction of complex marginal deformations of supergravity solutions to a class of general type IIB backgrounds that
include the gravity duals of N = 4 gauge theories at finite temperature and R-charge chemical
potentials and at the Coulomb branch. For these theories, we have concentrated on three simple
cases of the general solution, we have presented in full detail the marginally-deformed metrics,
and we have checked that their thermodynamics (for the rotating case) are the same as for the
undeformed ones, as they should.
Having constructed the marginally-deformed spacetimes, we considered their Penrose limits
for the multicenter case along a certain class of geodesics inside the angular part of the geometry.
Besides recovering familiar results, we have extended them to take account of the presence of
-deformations, which has not been considered in the literature up to date (with the exception
of the archetypal example of [3]), as well as for the presence of rotation (previously examined
in [27] for the undeformed solutions). We have also considered Penrose limits along a non-BPS
geodesic.
We next turned to a study of the giant gravitons supported on the PP-wave spacetimes just
constructed, using the simple example of the PP-wave of [3] but taking the -deformation into
account. For that case we found that, unlike -deformations, -deformations lift the degeneracy
between the giant and the point graviton but, nevertheless, the former remains stable under small
perturbations. We also showed that, for the particular geodesic under consideration, this result
is unaffected by the presence of rotation. We also considered dual giant gravitons, in which
case the situation remains qualitatively unchanged by the deformation. It would be interesting to
generalize these studies to more complicated PP-waves and to seek giant graviton solutions in
the full -deformed geometry.
Finally, we considered the standard Wilson-loop calculation of interquark potentials and
screening lengths for the dual gauge theory. Here we have considered the static heavy-quark
potential in the case of the Coulomb branch of the -deformed gauge theory, and for the two
D3-brane distributions under consideration we found a linear confining potential and a screened
Coulombic potential respectively. For the first case, we elucidated on the nature of the transition
to the confining phase and we calculated the critical value of the deformation parameter, while
for the second case we demonstrated invariance of the screening length under the deformation.
As noted in the text, it is very interesting to examine the stability of the string trajectories used to
calculate potentials in the deformed theory using the formalism developed in [46]. Also it is of
some interest to extend these results to the theory at finite temperature and chemical potential.
Acknowledgements
K.S. acknowledges support provided through the European Communitys program Constituents, Fundamental Forces and Symmetries of the Universe with contract MRTN-CT-2004005104, the INTAS contract 03-51-6346 Strings, branes and higher-spin gauge fields and the
Greek Ministry of Education programs YAOPA with contract 89194.
Appendix A. T- and S-duality rules
Here, we state our conventions for the T- and S-duality transformations used in Section 2.
Starting from T-duality, we consider a type II configuration characterized by the metric GMN ,
95
the NSNS 2-form B2 , the dilaton and the RR p-forms Ap and we wish to T-dualize along an
isometry direction, say y. Splitting the coordinates as x M = (x , y), we decompose the metric,
the NSNS 2-form and the RR p-forms into the quantities
a1 Gy dx ,
Gyy ,
1
b2 B dx dx ,
2
b1 By dx ,
(A.1)
and
1
A... dx dx dx ,
p!
1
p1
A...y dx dx ,
(p 1)!
(A.2)
so that, in particular, B2 = b2 b1 dy and Ap = p + p1 dy. Under T-duality, the NSNS

fields transform among themselves by the usual Buscher rules [48] while the RR fields transform
by terms involving both NSNS and RR fields [49]. In our present notation, these transformations
can be written in the compact form

MN dx M dx N = G dx dx + 1 (dy + b1 )2 a12 ,
G
B 2 = b2 1 a1 (dy + b1 ),
e2 = 1 e2 ,

A p = p + p1 1 p2 a1 (dy + b1 ).
(A.3)
This form of the T-duality rules is particularly useful in the computations of Section 2, as it
allows us to perform the transformations directly in form notation.
We next consider an S-duality transformation of a type IIB configuration, which acts on the
axion-dilaton = A0 + ie and the NSNS and RR 2-forms as follows

1 B2
B2
a + b
a b
,
=
(A.4)
= T
, =
,
c d
c + d
A2
A 2
leaving the Einstein-frame metric and the RR 5-form field strength invariant. For the purpose of
-deformations, we consider the particular SL(2, R) element

1 0
=
,
(A.5)
1
which, in addition, leaves the RR 2-form potential A2 invariant. The resulting transformations of
all fields, including the string-frame metric and the RR 4-form, are written explicitly as
MN = 1/2 GMN ,
G
B 2 = B2 A2 ,
e2 = 2 e2 ,

A 0 = 1 A0 (1 A0 ) e2 ,
1
A 4 = A4 A2 A2 ,
2
where
(1 A0 )2 + 2 e2 .
(A.6)
(A.7)
96
References
[1] J.M. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200;
S.S. Gubser, I.R. Klebanov, A.M. Polyakov, Phys. Lett. B 428 (1998) 105, hep-th/9802109;
E. Witten, Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150;
E. Witten, Adv. Theor. Math. Phys. 2 (1998) 505, hep-th/9803131.
[2] R.G. Leigh, M.J. Strassler, Nucl. Phys. B 447 (1995) 95, hep-th/9503121.
[3] O. Lunin, J. Maldacena, JHEP 0505 (2005) 033, hep-th/0502086.
[4] S. Frolov, JHEP 0505 (2005) 069, hep-th/0503201.
[5] S.A. Frolov, R. Roiban, A.A. Tseytlin, Nucl. Phys. B 731 (2005) 1, hep-th/0507021.
[6] R.C. Rashkov, K.S. Viswanathan, Y. Yang, Phys. Rev. D 72 (2005) 106008, hep-th/0509058.
[7] C.h. Ahn, J.F. Vzquez-Poritz, JHEP 0507 (2005) 032, hep-th/0505168;
C.h. Ahn, J.F. Vzquez-Poritz, Class. Quantum Grav. 23 (2006) 3619, hep-th/0508075;
J.P. Gauntlett, S. Lee, T. Mateos, D. Waldram, JHEP 05030 (2005) 0508, hep-th/0505207.
[8] S.A. Frolov, R. Roiban, A.A. Tseytlin, JHEP 0507 (2005) 045, hep-th/0503192;
S.A. Frolov, R. Roiban, A.A. Tseytlin, Nucl. Phys. B 731 (2005) 1, hep-th/0507021;
U. Grsoy, C. Nez, Nucl. Phys. B 725 (2005) 45, hep-th/0505100;
N.P. Bobev, H. Dimov, R.C. Rashkov, Semiclassical strings in LuninMaldacena background, hep-th/0506063;
J.G. Russo, JHEP 0509 (2005) 031, hep-th/0508125;
H.Y. Chen, S. Prem Kumar, JHEP 0603 (2006) 051, hep-th/0511164;
L.F. Alday, G. Arutyunov, S. Frolov, JHEP 0606 (2006) 018, hep-th/0512253;
U. Grsoy, JHEP 0605 (2006) 014, hep-th/0602215;
C.S. Chu, V.V. Khoze, JHEP 0607 (2006) 011, hep-th/0603207;
R. Minasian, M. Petrini, A. Zaffaroni, JHEP 0612 (2006) 055, hep-th/0606257.
[9] S.S. Pal, Phys. Rev. D 72 (2005) 065006, hep-th/0505257;
D.Z. Freedman, U. Grsoy, JHEP 0511 (2005) 042, hep-th/0506128.
[10] D. Berenstein, R.G. Leigh, JHEP 0001 (2000) 038, hep-th/0001055;
D. Berenstein, V. Jejjala, R.G. Leigh, Nucl. Phys. B 589 (2000) 196, hep-th/0005087;
D. Berenstein, V. Jejjala, R.G. Leigh, Phys. Lett. 493 (2000) 162, hep-th/0006168.
[11] N. Dorey, JHEP 0408 (2004) 043, hep-th/0310117;
N. Dorey, T.J. Hollowood, JHEP 0506 (2005) 036, hep-th/0411163.
[12] A. Mauri, S. Penati, A. Santambrogio, D. Zanon, JHEP 0511 (2005) 024, hep-th/0507282;
F. Elmetti, A. Mauri, S. Penati, A. Santambrogio, JHEP 0701 (2007) 026, hep-th/0606125.
[13] P. Kraus, F. Larsen, S.P. Trivedi, JHEP 9903 (1999) 003, hep-th/9811120.
[14] R. Hernndez, K. Sfetsos, D. Zoakos, JHEP 0603 (2006) 069, hep-th/0510132;
R. Hernndez, K. Sfetsos, D. Zoakos, Fortschr. Phys. 54 (2006) 407, hep-th/0512158.
[15] J.M. Maldacena, Phys. Rev. Lett. 80 (1998) 4859, hep-th/9803002;
S.J. Rey, J.T. Yee, Eur. Phys. J. C 22 (2001) 379, hep-th/9803001.
[16] C. Ahn, J.F. Vzquez-Poritz, JHEP 0606 (2006) 061, hep-th/0603142.
[17] M. Cvetic, D. Youm, Nucl. Phys. B 477 (1996) 449, hep-th/9605051.
[18] J.G. Russo, K. Sfetsos, Adv. Theor. Math. Phys. 3 (1999) 131, hep-th/9901056.
[19] R. Penrose, Any SpaceTime Has a Plane-Wave as a Limit, Differential Geometry and Relativity, Reidel, Dordrecht,
1976.
[20] K. Sfetsos, Phys. Lett. B 324 (1994) 335, hep-th/9311010;
K. Sfetsos, A.A. Tseytlin, Nucl. Phys. B 427 (1994) 245, hep-th/9404063.
[21] R. Gven, Phys. Lett. B 482 (2000) 255, hep-th/0005061.
[22] M. Blau, J. Figueroa-OFarrill, C. Hull, G. Papadopoulos, JHEP 0201 (2002) 047, hep-th/0110242.
[23] D. Berenstein, J.M. Maldacena, H. Nastase, JHEP 0204 (2002) 013, hep-th/0202021.
[24] D. Sadri, M.M. Sheikh-Jabbari, Rev. Mod. Phys. 76 (2004) 853, hep-th/0310119.
[25] V. Niarchos, N. Prezas, JHEP 0306 (2003) 015, hep-th/0212111.
[26] R. de Mello Koch, J. Murugan, J. Smolic, M. Smolic, JHEP 0508 (2005) 072, hep-th/0505227;
T. Mateos, JHEP 0508 (2005) 026, hep-th/0505243.
[27] A. Brandhuber, K. Sfetsos, JHEP 0212 (2002) 050, hep-th/0212056.
[28] M. Blau, M. OLoughlin, Nucl. Phys. B 654 (2003) 135, hep-th/0212135;
M. Blau, M. OLoughlin, G. Papadopoulos, A.A. Tseytlin, Nucl. Phys. B 673 (2003) 57, hep-th/0304198.
[29] J. McGreevy, L. Susskind, N. Toumbas, JHEP 0006 (2000) 008, hep-th/0003075.
97
[30] R.C. Myers, JHEP 9912 (1999) 022, hep-th/9910053.

[31] M.T. Grisaru, R.C. Myers, O. Tafjord, JHEP 0008 (2000) 040, hep-th/0008015;
A. Hashimoto, S. Hirano, N. Itzhaki, JHEP 0008 (2000) 051, hep-th/0008016.
[32] S.R. Das, A. Jevicki, S.D. Mathur, Phys. Rev. D 63 (2001) 044001, hep-th/0008088;
S.R. Das, S.P. Trivedi, S. Vaidya, JHEP 0010 (2000) 037, hep-th/0008203;
A. Mikhailov, JHEP 0011 (2000) 027, hep-th/0010206;
R.C. Myers, O. Tafjord, JHEP 0111 (2001) 009, hep-th/0109127;
I. Bena, D.J. Smith, Phys. Rev. D 71 (2005) 025005, hep-th/0401173.
[33] V. Balasubramanian, M. Berkooz, A. Naqvi, M.J. Strassler, JHEP 0204 (2002) 034, hep-th/0107119;
S. Corley, A. Jevicki, S. Ramgoolam, Adv. Theor. Math. Phys. 5 (2002) 809, hep-th/0111222;
D. Berenstein, Nucl. Phys. B 675 (2003) 179, hep-th/0306090;
R. de Mello Koch, R. Gwyn, JHEP 0411 (2004) 081, hep-th/0410236.
[34] S. Prokushkin, M.M. Sheikh-Jabbari, JHEP 0407 (2004) 077, hep-th/0406053.
[35] R. de Mello Koch, N. Ives, J. Smolic, M. Smolic, Phys. Rev. D 73 (2006) 064007, hep-th/0509007.
[36] M. Pirrone, JHEP 0612 (2006) 064, hep-th/0609173.
[37] E. Imeroni, A. Naqvi, JHEP 0703 (2007) 034, hep-th/0612032.
[38] K. Skenderis, M. Taylor, JHEP 0206 (2002) 025, hep-th/0204054;
H. Takayanagi, T. Takayanagi, JHEP 0212 (2002) 018, hep-th/0209160.
[39] A. Hamilton, J. Murugan, Giant gravitons on deformed PP-waves, hep-th/0609135.
[40] S.R. Das, A. Jevicki, S.D. Mathur, Phys. Rev. D 63 (2001) 024013, hep-th/0009019.
[41] S.J. Rey, S. Theisen, J.T. Yee, Nucl. Phys. B 527 (1998) 171, hep-th/9803135;
A. Brandhuber, N. Itzhaki, J. Sonnenschein, S. Yankielowicz, Phys. Lett. B 434 (1998) 36, hep-th/9803137;
A. Brandhuber, N. Itzhaki, J. Sonnenschein, S. Yankielowicz, JHEP 9806 (1998) 001, hep-th/9803263.
[42] A. Brandhuber, K. Sfetsos, Adv. Theor. Math. Phys. 3 (1999) 851, hep-th/9906201.
[43] I.S. Gradshteyn, I.M. Ryzhik, Table of Integrals, Series and Products, fifth ed., Academic Press, New York, 1994.
[44] P. Byrd, M. Friedman, Handbook of Elliptic Integrals for Engineers and Physicists, second ed., Springer, Heidelberg,
1971.
[45] H.B. Callen, Thermodynamics and Introduction to Thermostatistics, second ed., John Wiley & Sons, New York,
1985.
[46] S.D. Avramis, K. Sfetsos, K. Siampos, Nucl. Phys. B 769 (2007) 44, hep-th/0612139.
[47] S.D. Avramis, K. Sfetsos, K. Siampos, Stability of string configurations dual to quarkonium states in AdS/CFT,
hep-th/0706.2655.
[48] T. Buscher, Phys. Lett. B 194 (1987) 59;
T. Buscher, Phys. Lett. B 201 (1988) 466.
[49] E. Bergshoeff, C.M. Hull, T. Ortn, Nucl. Phys. B 451 (1995) 547, hep-th/9504081.
Chiral symmetry breaking as open string tachyon

condensation
Roberto Casero a , Elias Kiritsis a,b , ngel Paredes a,
a CPHT, cole Polytechnique, UMR du CNRS 7644, 91128 Palaiseau, France
b Department of Physics, University of Crete, 71003 Heraklion, Greece
Received 4 April 2007; accepted 4 July 2007

Abstract
We consider a general framework to study holographically the dynamics of fundamental quarks in a
confining gauge theory. Flavors are introduced by placing a set of (coincident) branes and antibranes on a
background dual to a confining color theory. The spectrum contains an open string tachyon and its condensation describes the U (Nf )L U (Nf )R U (Nf )V symmetry breaking. By studying worldvolume gauge
transformations of the flavor brane action, we obtain the QCD global anomalies and an IR condition that
allows to fix the quark condensate in terms of the quark mass. We find the expected Nf2 Goldstone bosons
(for mq = 0), the Gell-MannOakesRenner relation (for mq small) and the mass. Remarkably, the linear
confinement behavior for the masses of highly excited spin-1 mesons, m2n n is naturally reproduced.
Keywords: Gauge-gravity correspondence; Tachyon condensation; QCD; Chiral symmetry breaking
1. Introduction
The original AdS/CFT conjecture [1,2] states the existence of a stringy description of a
conformal, highly symmetric, quantum field theory. The low energy limit on the string side (supergravity) corresponds to the strong coupling regime of the gauge theory. A natural question that
has attracted a lot of attention is whether such a duality can be generalized in order to describe
the strong coupling regime of more realistic gauge theories. In the last few years we have learned
E-mail address: paredes@cpht.polytechnique.fr (. Paredes).

doi:10.1016/j.nuclphysb.2007.07.009
R. Casero et al. / Nuclear Physics B 787 (2007) 98134
99
that the ideas of holography can indeed be applied to a certain extent in that direction. In particular, higher-dimensional duals of (non-supersymmetric) QCD-like theories have been built and
many qualitative and, surprisingly, some quantitative features of real-world strong interactions
can be described in such frameworks with reasonable accuracy.
One of the most important strong-coupling phenomena of QCD is spontaneous chiral symmetry breaking U (Nf )L U (Nf )R U (Nf )V by the formation of a non-perturbative quark
condensate qq
= 0 (Nf denotes the number of quark flavors). The goal of this paper is the
study of this symmetry breaking using holographic techniques.
We start by briefly reviewing the results obtained on this topic in the existing literature.
This question was initially addressed in [3,4] (see also [5]), by considering stacks of Nf flavor
branes in non-supersymmetric backgrounds created by Nc color branes. The chiral symmetry is a
U (1)A isometry of the geometry and is spontaneously broken due to the embedding of the flavor
branes. Typically, these models are asymptotically supersymmetric in the UV. This allows to have
good control on quantities like the quark condensate and the quark mass. Imposing regularity of
the brane configuration, one obtains an IR condition which fixes the quark condensate in terms
of the quark mass, as expected in QCD. Also, by introducing a small quark mass, the Gell-Mann,
Oakes, Renner (GOR) relation for the pion masses
m2 = 2
mq qq
,
f2
N
mq 0
can be obtained. The O( Nfc ) mass of the in this kind of setup was studied in [6]. The limitation
of this approach is that, even for massless quarks, only the Abelian chiral symmetry is present,
while the non-Abelian chiral subgroup of the flavor symmetry is absent from the beginning, and
its breaking cannot therefore be seen. From the field theory side, this is due to the existence of
a qq
Yukawa coupling (color and flavor branes are codimension four), where is one of the
(massive) scalars living on the brane world-volume [4].
A different approach was followed in [7,8]. By considering Nf D8D8 pairs in a nonsupersymmetric background corresponding to Nc D4 branes, a U (Nf ) U (Nf ) global symmetry was introduced. Contrary to the case described earlier, the chiral symmetry is now realized on
the flavor brane world-volume rather than in the geometry. The breaking of the full non-Abelian
chiral symmetry is due to a smooth recombination of branes and antibranes in the IR. By computing the meson spectrum, Nf2 massless Goldstone bosons are found. One of them, the , is
N
massless only when Nc and the authors addressed the problem of finding its O( Nfc ) mass.
The chiral anomaly and associated WZW term are nicely reproduced. Remarkably, quantities
(some meson masses, decay constants and couplings) computed in this holographic setup match
reasonably well the experimental values. Moreover, the model also incorporates a very natural
picture of chiral symmetry restoration at high temperature: once a horizon is formed in the geometry, the branes and antibranes can fall into it and they do not need to recombine [9].
This construction however presents some shortcomings too. First, there is no parameter which
can be associated to the quark bare masses. By modifying the flavor brane embedding, one can
vary the masses of the massive mesons, which can be thought of as a modification of the constituent quark masses. However, the pions remain massless, implying that the bare quark masses
are always vanishing (see [912]). This is clearly not a physical feature of QCD. Related to this
problem, there is no parameter that can be identified with the quark condensate, even though this
is a crucial quantity to describe spontaneous chiral symmetry breaking in QCD. Another unphys-
100
ical feature of the model is the absence of a tower of massive mesons with the quantum numbers
of the pion.
Finally, there is an interesting alternative approach, that has been named AdS/QCD [13,14].
It is a bottomup construction in the sense that it does not emerge directly from some known
string theory. Rather, it just assumes the existence of a dual description of QCD living in a fivedimensional, asymptotically AdS space. The five-dimensional bulk fields needed to dualize the
quark bilinears are introduced by hand, and spontaneous chiral symmetry breaking is modeled by
giving a vev to a bulk scalar by hand. The model incorporates the existence of Goldstone bosons,
vector meson dominance [14] and the GOR relation for the pion masses [13]. In its simplest form,
the five-dimensional space is taken to be just AdS5 with a hard IR cutoff. However, modifications
of the background have been considered in order to incorporate the effect in the geometry of
the possible condensates [15] or linear confinement [16]. A drawback of the model is that, since
the condensate is not determined from some dynamical computation, there is an extra parameter
compared to QCD. Also, being ad hoc, the models are less satisfying and it would be nice to
understand how they are related to some (possibly non-critical) string theory which may help
fix the background geometry or the potential for the bulk scalar. Some progress along this line
seems imminent, [17].
The goal of this paper is to improve the present holographic description of chiral symmetry breaking by proposing a general string theory construction which accounts for all relevant
phenomena in a natural and simple way. Flavors are introduced via a stack of Nf overlapping
braneantibrane pairs. At the same time, we allow for a non-trivial profile for the lightest open
string mode which transforms in the bifundamental representation of the U (Nf )L U (Nf )R
flavor symmetry group supported by the braneantibrane pairs. In flat space, this mode would
be an open string tachyon and we will refer to it as such in our framework. Nevertheless, we
emphasize that it does not generate any instability in the setup. This complex scalar is the natural candidate to describe chiral symmetry breaking, as suggested in [7] (see also [11,18,19]).
A review on open string tachyon physics can be found in [20].
We will study the world-volume effective action for the braneantibrane system, including
the tachyon both in the DiracBornInfeld and WessZumino terms. To keep the focus on the
relevant dynamics, we will consider a simple setup in which only the tachyon and the worldvolume gauge fields of the branes and antibranes are dynamical. This is the minimal setup to
include the quark bilinears in the dual theory (up to spin one) and it matches the bulk field
content of the AdS/QCD models [13,14]. We will also not worry about the closed string physics.
We will keep the discussion very general, without applying it to any particular model: we will
only need to assume that the background metric is such that the color theory is confining (using
a general prescription in [21]).
In a few places we will also perform some more explicit computations; in these cases we
will assume that the space is asymptotically AdS in the UV. This will allow us to apply the
general formalism developed in this paper to a relevant and interesting example, while hopefully
clarifying our presentation.
1.1. Summary of results
The open string tachyon we introduce transforms in the bifundamental representation of the
flavor symmetry group U (Nf )L U (Nf )R , and couples on the boundary to the quark scalar and
pseudoscalar bilinears. Up to a normalization we will introduce later on, we have, in fact:
T q
1+5
q,
2
T q
15
q.
2
101
(1.1)
By considering braneantibrane configurations, and including the dynamics of the complex

open string tachyon scalar T , we have been able to provide a general framework to describe
holographically the chiral dynamics of QCD and related theories, and to reproduce many features
of QCD:
The UV non-normalizable component of the tachyon corresponds to the quarks bare (complex) mass matrix mq . For simplicity, in this paper we will consider all quarks to have the same
(possibly vanishing) mass mq .
The UV normalizable mode of the tachyon corresponds, in turn, to turning on an expectation value for the quark condensate.
We show that in a confining background, even for mq = 0, the tachyon profile cannot be
trivial, which results in the chiral part U (Nf )A of the flavor symmetry being always broken, either explicitly or spontaneously. The construction of this paper, therefore, provides a holographic
version of the ColemanWitten theorem [23].
The fact that the tachyon diverges in the IR (signalling the IR fusion of branes and antibranes) constrains the way it vanishes in the UV: this allows to fix the value of the quark
condensate qq
in terms of the quark bare mass mq .
We can derive formulae for the anomalous divergences of flavor currents, when they are
coupled to an external source.
The WZ part of the flavor brane action gives the AdlerBellJackiw U (1)A axial anomN
aly [24], and an associated Stuckelberg mechanism gives an O Nfc mass to the would-be
Goldstone boson , in accordance with the VenezianoWitten formula.
When mq = 0 (i.e. the tachyon has no UV non-normalizable mode), we find that (in the
Nc limit) the meson spectrum contains Nf2 massless pseudoscalars, the U (Nf )A Goldstone
bosons.
Studying the spectrum of highly excited spin-1 mesons, we find the expected property
of linear confinement: m2n n. However, the slopes we obtain for the vector and axial vector
trajectories are different, in contrast with certain expectations (see for instance [22]). This may
be due to the fact that it is not clear whether the DBI action we use to include the effects of
the dynamics of the open string tachyon remains valid in the region where the tachyon and its
derivative diverge.
When we consider an asymptotically AdS space in the UV, we can precisely reproduce the
GOR relation for the mass of the pions in the case of a small but non-zero quark bare mass.
On the other hand, the main limitations our approach still presents are the following:
The details of the tachyon potential that one should insert in the DBI action are not rigorously known. Despite this, its gross features suffice for extracting qualitative information.
A background in which the braneantibranes are on top of each other (and with no Wilson
line turned on), would probably require large curvature, as we explain. Then, a reliable background metric cannot be found by solving Einsteins equations.
However we found encouraging that a large amount of expected QCD features can be extracted from our construction despite the misgivings above.
102
2. The general construction

The general setup we consider consists of a system of Nf overlapping DpDp flavor brane
antibrane pairs in the background associated to Nc Dq color branes. In critical, ten-dimensional
string theory, we will need to impose the condition p + q = 12 and require the color and flavor
branes to be codimension 6, in order to get the correct WZ couplings.1 For instance, one can
think of a D3D9D9 intersection as in [18] or a D4D8D8 intersection similar to the Sakai
Sugimoto model [7]. Notice, though, that the general formalism we develop can also be applied
to lower dimensional setups provided the WZ couplings are the obvious adaptation of those of
Section 3. Interesting examples of this kind are the AdS/QCD models or non-critical string constructions (N = 1 theories with spacetime filling braneantibrane flavors have been discussed
in [19,25,26] and N = 0 in [10,19]).
In the spectrum of stacks of overlapping braneantibrane pairs there is a tachyonic, complex scalar, open string mode, which transforms in the bifundamental representation of the
U (Nf )L U (Nf )R flavor symmetry group supported by the braneantibrane pairs. As we argued in the introduction, we can in general associate chiral symmetry breaking to a vacuum
expectation value for this field. Near the UV region (in our coordinates this is around z = 0),
the tachyons vev vanishes and the full U (Nf )L U (Nf )R is present, as in the Lagrangian of
massless QCD. As we approach the IR, we prove in Section 3.2 that if the theory is confining the
tachyon must acquire a non-trivial vev, breaking the symmetry to its diagonal U (Nf )V subgroup,
(see also Section 2.2). Concretely, we show that the vev reaches infinity at a finite distance in the
bulk, a point that we denote by zIR . In a sense, this process can be thought of as braneantibrane
recombination. In the minimal construction that we consider in the rest of this paper, zIR coincides with the end of space. Fig. 1 presents a general profile for the tachyon in the setup we just
described.
Fig. 1. A scheme of the setup. Generically, the tachyon vanishes in the UV where one has a stack of braneantibrane
pairs. In the IR, it diverges and the braneantibrane pairs end smoothly. Here stands for the modulus of the tachyon
field.
1 The reason why is very simple to explain. In Section 3.2 we will be looking for a coupling on the flavor DpDp branes
which involves 6-dimensional gauge field forms. This can only come from the WessZumino coupling of the p-branes to
a magnetic k-form potential, where k = p + 1 6 = p 5. Now Dq-branes source a Cq+1 RR potential, which is parallel
to the source branes and generically depends on the radial direction alone. Imposing 4-dimensional Lorentz invariance of
the background ensures that the Hodge dual of Cq+1 is a magnetic k-form,
with k = 10 (q + 2) 1 = 7 q. Imposing
k = k finally gives the condition p + q = 12.
103
For the present discussion, it will be very important to consider the completion of the worldvolume action of the braneantibrane system to include also the couplings to the open string
tachyon. As usual, this action is composed of a DiracBornInfeld and a WessZumino term.
In the rest of this section and in Section 4, we study the physics of the DBI piece, which yields
the vacuum of the configuration and the meson mass spectrum. The WZ part will be studied in
Section 3. From it, we derive the holographic description of the parity and charge conjugation
symmetries, the global anomalies, and a holographic ColemanWitten theorem.
2.1. The degrees of freedom: A minimal setup
As proposed in [27,28], the semiclassical action for a DpDp system which includes the
physics of the open string tachyon is a generalization of the usual DiracBornInfeld action for
D-branes.
We start from the simpler configuration of a single braneantibrane pair. In this case the
proposed DBI action reads:

S = d p+1 x e V 2 , YLI YRI , x
(2.1)
det AL + det AR
where (we will set 2 = 1 from now on, and only reinsert when needed):
(i)
1
1
(DM T ) (DN T ) + (DN T ) (DM T ),

L
DM T = M + iAM iAR
M T
A(i)MN = P [g + B]MN + FMN +

(i)
(i)
(i)
FMN = M AN N AM ,
(2.2)
and V ( 2 , YLI YRI , x) is the tachyon potential. The complex tachyon is denoted by T ei ,
I are the transverse scalars (which enter
the indices i = L, R denote the brane or antibrane, the Y(i)
the expression for A(i)MN through the pullback), and A(i) is the world-volume gauge field. Here
and in the following, capital Latin characters, M, N , denote world-volume directions, which
include in particular the Minkowski directions, which we will denote by Greek letters , , and
the holographic radius which will be referred to as z.
For simplicity, we will not consider a background B-field. We will also ignore the transverse
scalars living on the flavor branes. These modes, which are generally present in a critical string
setup, do not have any obvious QCD interpretation and would not play any relevant role in the
present discussion. Thus, we only consider the degrees of freedom coming from the open string
tachyon and the world-volume gauge fields along the Minkowski or the holographic directions
(i)
(i)
A , Az . This field content is enough to include the basic QCD quark bilinears up to spin 1 and
coincides with the one of the AdS/QCD models.
Higher spin operators correspond to stringy excitations and cannot be addressed in this formalism. They could be introduced by hand as in [29].
Since we are ignoring transverse scalars, for the tachyon potential we take the expression for
overlapping braneantibranes2 derived in boundary string field theory [30,31]3 :
2 If one considered more complicated setups with non-trivial brane profiles, the potential should depend also on the
braneantibrane distance. For a discussion in a related setup, see [11].
3 Unfortunately, there are not unified conventions in the literature about the definition of T . The unambiguous statement
1 . Eq. (2.3) is thus consistent with (2.1)
is that the mass squared of the open string tachyon in flat space is m2T = 2
104

2
V 2 = Tp e .
(2.3)
One should keep in mind that this expression is not rigorously justified since there can be nontrivial field redefinitions between the two formalisms. Moreover the boundary string field theory
computation is performed in flat space and it is not trivial at all that the result should apply
unchanged to curved backgrounds. Remarkably, in Section 4.2 we will find the nice feature
of linear confinement using this Gaussian behavior for the potential. Most of the rest of the
discussion is insensitive to the precise expression of the tachyon potential.
After presenting the simpler Abelian case, we can now go to considering a stack of Nf > 1
braneantibrane pairs, which is what we are eventually interested in. The natural non-Abelian
generalization of (2.1) is:

S = d p+1 x Sym Tr e V T T
(2.4)
det AL + det AR
where Sym Tr is a symmetric trace as defined in [32]. The analogous proposal for the action of a
stack of non-BPS D-branes was made in [33] and can be related to systems of branesantibranes
along the lines of [28]. Although the symmetric trace prescription for the non-Abelian action is
known to be corrected at higher orders in , this will not be relevant for our discussion. All that
is important is that the potential appears in a single trace, and this is guaranteed by the large Nc
limit, [23].
The gauge group supported by the braneantibrane system is U (Nf )L U (Nf )R . The gauge
fields AL,R transform in the adjoint representation of U (Nf )L and U (Nf )R respectively, while
the tachyon T transforms in the bifundamental of U (Nf )L U (Nf )R , (that is, in the antifundamental of U (Nf )L and in the fundamental of U (Nf )R ). Its hermitian conjugate T transforms
in the opposite way. For more details on our choice of conventions, we refer the reader to
Appendix A.
2.2. The vev for the tachyon field
In this section we want to determine the vacuum configuration for a stack of braneantibranes
over a confining vacuum. It is trivial to show that the equations of motion allow for the gauge
fields to be consistently set to zero, AL = AR = 0. As we mentioned in the introduction,
and we will explicitly show in Section 3.2, anomaly considerations regarding the WZ part of the
action of a stack of braneantibranes over a confining vacuum require that |zIR . This
necessitates that we allow for a non-trivial profile of the tachyon in the vacuum configuration.
Four-dimensional Poincar invariance determines the general form of the bulk five-dimensional metric. In the mostly plus signature this reads:
ds52 = gxx (z) dx dx + gzz (z) dz2 .
(2.5)
The non-Abelian action (2.4) is a very complicated object, where different components of the
fields mix because of non-trivial commutation relations. We start from the simpler Nf = 1 case,
where these complications do not arise. From (2.1), it is consistent to set the tachyon phase to a
constant.

1 2 ), where we have considered flat
and (2.2) since, then, for small the action is proportional to S ( 12 ( )2 12 2

space and reinserted .
105
We are thus left with the world-volume action for the modulus of the tachyon , which, upon
reducing to five dimensions reads:

2 2
2
4
eff
V gxx gzz + (z )2 .
S = 2 d x dz e
(2.6)
The prefactor eeff (z) is defined as the dilaton exponential times whatever comes from integrating the transverse, spectator, p 4 dimensions. Schematically:

2
gzz + (z )2
eeff (z)V 2 (z) gxx

d p4 y e V 2 , YLI YRI , x
det AL + det AR
(2.7)
N p3
N p3 is the
internal part of the flavor branes world-volume.4

where
The EulerLagrange equation for obtained from (2.6) reads:

2 eeff )
2 eeff )
2 z (gxx
z (gxx
z gzz
V
3
2
(
)
+
+
(
)
z2 +
g
z
z
zz
z
2 eeff
2 eeff
gzz gxx
2gzz
V
2
gxx
= 0.
(2.8)
In the following we will use the potential (2.3), so VV = 2 . Eq. (2.8) is a second order differential equation which therefore has two integration constants. Since is dual to the
quark bilinear, we know on general AdS/CFT grounds that, by looking at the UV behavior
of , these two constants can be related to the quark bare mass and the quark condensate. We
2 /z2 + and
now assume that the space is asymptotically AdS, that is gxx (z) gzz (z) RAdS
e eff const e 0 for small z. The tachyon field is dual to the quark bilinear which has dimension 3 in weak coupling perturbation theory. Thus, the usual relation between the mass of a five2 ,
dimensional scalar field and the conformal dimension of its dual operator, ( 4) = m2 RAdS
2 = 3.5 Assuming that the tachyon mass takes the same value as in flat space6
fixes m2T RAdS
1
2
2 = 6 = 3 . This leads to the followmT = 2 = , the radius of AdS should be small: RAdS
ing UV behavior (in order to simplify notation the vacuum expectation value will be denoted
just as from now on):

1
3
3
2
4Tp
2
2
RAdS can RAdS 2
0
mT e
3

= mq (z + ) + z + (small z).
(2.9)
4 The reader might be puzzled by the approximation involved in discarding all dependence on the internal world-
volume scalars. It should be noted, though, that our ansatz for the five-dimensional metric (2.5) is completely general,
being constrained only by four-dimensional Poincar invariance. The vacuum configuration of the internal scalars will
only affect the explicit form of the metric coefficients and the radial dependence of the tachyon modulus and dilaton.
Since our arguments are completely independent of the particular form of the confining, reduced, five-dimensional background, all the following results apply regardless what the critical string setup is.
5 It is possible for some theories that are asymptotically conformal and strongly coupled that the dimension of the
quark bilinear departs from its free field value. This is the case for example in the Seiberg conformal window where such
Nc
quark bilinears have dimensions = 4 3 N
, and vary between 2 and 3 inside the conformal window. However this
f
example is not relevant here as we are working at Nf Nc .

6 The mass of the tachyon field could change in a curved background or, for instance, due to a non-trivial vev for a
Wilson line living on the flavor branes worldvolume. This could modify the details of the tachyon potential but we do
2 = 3 is satisfied.
not expect the generic features of the discussion to change as long the assumption m2T RAdS
106
We have written this expression in terms of the canonically normalized can , such that the
3/2
boundary coupling is RAdS d 4 x can z1 qq.
The parameter can be related to the quark condensate qq,
see Appendix B for details. The UV behavior (2.9) is the same as the one for the
scalar breaking the chiral symmetry in the AdS/QCD models [13,14].
We now want to prove that the IR condition |zIR gives a condition on (z), fixing ,
and therefore the condensate, in terms of the mass mq .
A general result of the gauge-gravity correspondence states that a sufficient condition for
confinement is that at some zdiv , gzz (zdiv ) while gxx (zdiv ) = 0 and z gxx (zdiv ) < 0 [21].
We will assume this is the case for our background and in particular, after possibly redefining z,
we fix the divergence to be a single pole gzz = b(zdiv z)1 near z = zdiv .
This is the behavior, for instance, in Wittens dual of YangMills [34]. Inspecting the leading
terms for large of (2.8), we see that the tachyon can only diverge at zIR = zdiv and there it
g 2 eeff
diverges as (zIR z)a , where a = b (gxx2 eeff ) z=z is a positive number since the
IR
z xx
derivative inside the bracket is typically negative.
We warn the reader that this result should be taken with a grain of salt because it relies on
the behavior of (2.8) when , , which lies outside the regime of validity of the DBI
action (2.6). Nonetheless, the general lesson one should learn from this computation is that the
|zIR consistency condition constrains the IR behavior of the tachyon. Physically, this fixes
the quark condensate in terms of the quark mass,7 in the spirit of [3,4].
To conclude the present discussion on the tachyon profile for the Abelian Nf = 1 case, we
should note that = 0 (which would leave the chiral symmetry unbroken) is the simplest solution
of (2.8). At first sight, one could be tempted to discard this solution on energetic grounds: since
the tachyon stays at the top of the potential, it will want to roll down. But this reasoning may be
naive. The (static) energy of the configuration is proportional to minus the on-shell action (2.6).
Thus, there are two competing contributions, a growing tachyon tends to lower the action through
V ( 2 ) but to increase it due to the derivative inside the square root. It is hard to tell which effect
dominates without calculating the Euclidean action in a specific background. We therefore stress
that the general argument to enforce an IR growing vev for the tachyon field in QCD-like theories
is not energetic, but rather relies on the anomaly-based condition |zIR , stated at the
beginning of this section and derived in Section 3.2. This discussion is closely related to the one
in [4] where a chiral-preserving embedding was discarded because it would correspond to an
open D6-brane.
At this point we can consider the more general non-Abelian case Nf > 1. Throughout the
paper, we will consider for simplicity that all Nf flavor fields have the same mass mq . The symmetrized trace in the action (2.4) might seem to make it a lot more complicated to find a vacuum
configuration for the tachyon matrix T . Fortunately we can show that there is a perturbatively stable configuration given by Nf identical copies of the solution for the Abelian case we described
above
T = (z)I.
(2.10)
In fact, we can apply here a result that was derived in [4] for a different setup. The reasoning
in [4] showing that the vacuum configuration where vevs are proportional to identity matrices,
like (2.10) in our case, is a stable (up to quadratic order) solution of the non-Abelian DBI action
7 It is natural that for a fixed m , the conditions stated above fix uniquely . However, within the present very general
q
formalism, this uniqueness cannot be proved. We will come back to this in Section 3.4.
107
is very general. It only relies on having a single trace in front of the action and the fact that the
difference between Abelian and non-Abelian DBI actions consists entirely of terms involving
commutators. The argument does not prove that this is the global minimum. There could in
principle be a different configuration with lower energy, but this is an unlikely possibility and we
will assume that (2.10) describes the vacuum of the theory.
Once we have established that T is proportional to the identity matrix, the equation of motion for (z) is the same as (2.8) as well as the UV and IR conditions, since for the configuration
(2.10) and up to the order we need here, the non-Abelian action (2.4) is actually the sum of Nf
identical copies of the Abelian action (2.1).
Looking at how the fields transform under flavor symmetry transformations (A.6), it is clear
that the vacuum (2.10) spontaneously8 breaks U (Nf )L U (Nf )R into its diagonal subgroup
U (Nf )V , where the vectorial gauge transformation is VL = VR .
When mq = 0, this spontaneous breaking implies the existence of Nf2 Goldstone bosons. We
will explicitly find them in the spectrum in Section 4.3 (see also Section 3.5 for an indirect
argument). We conclude this section by noting that the fact that the symmetry is broken down
to U (Nf )V , and not further, is also related to a theorem by Vafa and Witten [35] that states that
vectorial symmetries cannot be spontaneously broken.
3. The D-brane WessZumino sector
We now consider the WZ coupling of the Nf DpDp branes to the background RR fields,
including the tachyon dependence. This part of the action is given by (the following expression
was proposed in [36] and proved in [37,38] using boundary string field theory):

C Str exp[i2 F]
SWZ = Tp
(3.1)
p+1
where p+1 is the world-volume of the DpDp branes, C is a formal sum of the RR potentials

pn+1
C = n (i) 2 Cn , and F is the curvature of a superconnection A. In terms of the tachyon
field matrix T and the gauge fields AL and AR living respectively on the branes and antibranes,
they are (from now on, we will set 2 = 1 again and use the notation of [37]):

iAL T
iFL T T
DT
iA =
(3.2)
,
iF =
.
T
iAR
DT
iFR T T
In Appendix C we review the relevant definitions and properties of this supermatrix formalism
(in particular, notice the definition of the supermatrix product).
The superconnection is defined as:
F = dA iA A
(3.3)
8 From the field theory point of view the nature of the breaking depends on the asymptotic UV behavior of the tachyon.
When mq = 0, the vacuum configuration of the tachyon is not normalizable and we are adding a mass term for the
quarks to the Lagrangian: the breaking is explicit. If instead mq = 0, the tachyon is normalizable, and we are turning on
an expectation value for the quark bilinear in the original theory, which makes the symmetry breaking spontaneous. From
the gravity point of view, instead, both for vanishing or non-zero mq , the breaking is spontaneous since it is determined
by a non-trivial expectation value.
108
and satisfies the Bianchi identity:

dF iA F + iF A = 0.
(3.4)
Using this identity and the cyclic property of the supertrace (C.3) it is immediate to check that
Str ei F is a closed form and therefore that (3.1) is invariant under RR gauge transformations. At
least locally, there exists then a form such that:
d = Str ei F .
(3.5)
The case of interest here is the WZ coupling on the DpDp flavor branes in the background
of Nc Dq-color branes. As explained at the beginning of Section 2, we have p + q = 12 and the
electric RR potential sourced by the color branes is then Cq+1 = C13p , while its magnetic dual
is Cp5 . Formula (3.1) reduces to:

SWZ = iTp
Cp5 Str ei F 6-form
p+1
= i() Tp
p
Fp4 |5-form .
(3.6)
p+1
When the branes worldvolume p+1 has boundaries, the two expressions in (3.6) differ by
boundary terms, and are therefore not equivalent. In this case, which is the relevant one for
this paper, it is argued in [39] (for intersecting branes) that the correct form for the action is the
last one in (3.6). Since |5-form is defined only up to an exact form, there is then a possible ambiguity in the definition of the WZ action (3.6). We will assume that the correct form for |5-form
respects the discrete symmetries of Str ei F |6-form , which we will discuss in the following Section 3.1.
Notice that the potential sourced by the Dq-branes has indices parallel to the branes and
generically depends on the radial direction. Moreover, 4-dimensional Lorentz invariance of the
background (plus possibly a radial redefinition) prevents the metric from having non-zero offdiagonal terms in the directions (0, 1, 2, 3, z). These two facts ensure that Fp4 does not have
legs along the (0, 1, 2, 3, z) directions. Thus, the only components of the five-form |5-form 5
which we are interested in are those along x 0 , x 1 , x 2 , x 3 and z (we will call M5 this fivedimensional space). Similarly to Section 2, we assume at this point that 5 does not depend on
the directions along which Fp4 lies, as expected on physical grounds. Then we can integrate
out those directions in (3.6) to obtain the 5-dimensional effective ChernSimons action:

iNc
p
p
Fp4
5 = ()
5
SCS = i() Tp
(3.7)
(2)5 3
p+1 M
5
M5
where we have used the quantization condition

of the D-brane tensions.
1
2
2(10)
M5
Fp4 = Nc T12p and the explicit value
3.1. Discrete symmetries (P , C)

The six-form Str ei F |6-form appearing in (3.6) is invariant under two discrete symmetries,
P and C, that we will present in this subsection. Thus, an 5 invariant under P , C exists. It
follows then from (3.7) that the CS action is invariant under the same symmetries, which acquire
a fundamental physical interpretation in the dual field theory.
109
Parity
The six-form Str ei F |6-form is left invariant by the change P P1 P2 , where:
P1 :
AL AR ,
T T
P2 :
(x1 , x2 , x3 ) (x1 , x2 , x3 ).
and
Notice that P1 also transforms DT DT , FL FR . Indeed we can easily check that:

iFL T T
DT
Str ei F = Str exp
DT
iFR T T

P1
iFR T T
DT
Str exp
iFL T T
DT

iFL T T
DT
= Str exp
.
DT
iFR T T
(3.8)
(3.9)
For the last equality we have just swapped the ordering of the blocks inside the matrix, picking
up a minus sign due to the definition of the supertrace (C.2). The overall minus sign in (3.9) is
compensated by the minus sign acquired by the form dx1 dx2 dx3 under the action of P2 ,
and the CS action (3.7) is invariant under P .
We conclude that four-dimensional parity is equivalent to charge conjugation for the D-branes
(interchanging D D).
Charge conjugation
The other discrete symmetry of Str ei F |6-form , and consequently of the CS action (3.7), is:
t
C: AL AtR ,
(3.10)
AR AtL ,
T T t,
T T ,
where the superscript t denotes matrix transposition. Notice that under (3.10), the field strengths
and tachyon covariant derivatives transform as FL FRt , FR FLt , DT DT t , DT
(DT )t . The supertrace transforms as:

iFL T T
DT
iF
Str e = Str exp
DT
iFR T T

(DT )t
iFRt (T T )t
C
Str exp
DT t
iFLt (T T )t

iFLt (T T )t
DT t
= Str exp
(DT )t
iFRt (T T )t
pt

iFL T T
iDT
= Str exp
iDT
iFR T T

iFL T T
iDT
.
= Str exp
(3.11)
iDT
iFR T T
For the identity in the second line we have again swapped the ordering of the blocks picking up a
minus sign. In the equality of the following line we used the definition of pseudotranspose (C.4).
Finally in the last one, we used property (C.5) and the fact that the supertrace does not change
under pseudotransposition. The matrix of the last expression is not the same as the untransformed
one of the first line of (3.11). Therefore, (3.10) does not leave the supertrace invariant. However,
we are only interested in the 6-form within the full expression of the supertrace. Noticing that all
110
Table 1
A summary of the world-volume fields, their spins and their transformation properties under parity and charge conjugation. In the gauge we will use below (see Section 4.1 for more details), (AL AR )z and (AL + AR )z are set to zero
while the longitudinal component of (AL AR ) combines with the pseudoscalars
World-volume field
T +T
i(T T )
JPC
0++
0+
(AL +AR )
2
1
(AL AR )
2
1++
two-forms (FL , FR , DT DT ) inside the matrix of the last line have picked up a minus sign
compared to the initial expression, we find that the 6-form (in fact, all the 4k + 2-forms) is left
invariant.
If one wanted to maintain invariant the 4k-forms of the supertrace (which can be related
to gauge theories in 4k 2 dimensions in the same way as we relate the 6-form to a fourdimensional field theory), then the transformation would be AL AtL , AR AtR , T
(T )t , T T t . This is in agreement with the fact that charge conjugation in 4k 2 dimensions
does not change chirality.
Table 1 summarizes the transformation properties of the different degrees of freedom. In par
ticular, notice that the vacuum condensate (2.10) transforms as 0++ as expected for the qq
bilinear.
3.2. The chiral anomaly: External currents
We now want to study the anomaly of the chiral symmetry when the flavor currents are coupled to external sources, i.e. study the WZ term as an action for the flavor brane world-volume
gauge fields (AL , AR ) which couple in the boundary to combinations of the vector and axial currents, (see Table 1). As is usual for ChernSimons terms, a gauge transformation does not leave
the action (3.7) invariant but produces a boundary term. As in previous cases, [2], this term is
matched with the global anomaly of the dual field theory.
We set the tachyon to its vacuum value (2.10). It is a straightforward although lengthy computation to expand the supertrace in (3.6).

Str ei F 6-form

1
2
= Tr e iFL FL FL + iFR FR FR
6

1
1
+ 2i d (AL AR ) FL FL + FL FR + FR FL + FR FR
2
2
+ 2 (AL AR ) (AL AR ) (FL FL FR FR )
+ 2 (AL AR ) FR (AL AR ) FL
3 d (AL AR ) (AL AR ) (AL AR ) (FL + FR )
i
+ 4 (AL AR ) (AL AR ) (AL AR ) (AL AR ) (FL FR )
4

i 5
d (AL AR ) (AL AR ) (AL AR ) (AL AR ) (AL AR ) .
10
(3.12)
111
It is now not very hard to find a 5-form 5 such that d5 = Str ei F |6-form . We can write:

1
1
2
5 = Tr e iAL FL FL + AL AL AL FL
6
2
i
+ AL A L A L A L A L
10
1
i
+ iAR FR FR AR AR AR FR AR AR AR AR AR
2
10

i
2
+ iAL FR FR iAR FL FL + (AL AR ) (FL FR + FR FL )
2
1
1
+ AL AL AL FL AR AR AR FR
2
2

i
i
+ AL A L A L A L A L AR A R A R A R A R
10
10

+ i 3 d (AL AR AR AL ) (FL + FR ) + iAL AL AL AR

i
AL AR AL AR + iAL AR AR AR
2

i
+ 4 (AL AR ) (AL AR ) (AL AR ) (AL AR ) (AL AR ) .
20
(3.13)
We now consider a gauge transformation of (3.13). As we argued above, the tachyon vacuum
configuration (2.10) we are considering breaks the flavor invariance to the diagonal (vectorial)
subgroup of U (Nf )L U (Nf )R . Therefore, taking VL = VR = V , the gauge transformation of
the fields involved in (3.13) reads (see Appendix A for more details)
AL = iD = i d AL + AL ,
AR = iD = i d AR + AR ,
FL = [, FL ],
FR = [, FR ].
(3.14)
Inserting (3.14) in (3.13) we find:

1
2
5 = Tr e d 1 + 2 FL FL
6
i
(AL AL FL + AL FL AL + FL AL AL )
2
1
i
+ AL AL AL AL + FR FR + (AR AR FR
2
2

1
+ AR FR AR + FR AR AR ) AR AR AR AR
2

3
+ d (AL AR ) (FL + FR ) + (FL + FR ) (AL AR )
+ i(AL AL AL AL AL AR AR AL AL + AL AR AR

+ AR AR AL AR AR AR ) .
(3.15)
112
As expected, it is straightforward to show that the 5-form 5 is closed

d( 5 ) = 0.
(3.16)
The result (3.16) ensures that the gauge variation of the 5-dimensional CS action (3.7) is given
by 4-dimensional boundary terms. The UV boundary term (where 0) reproduces the field
theory global anomaly [2], as we will see below. However, if we consider a background dual to
a confining gauge theory, the gauge variation of the CS action (3.7) may receive a contribution
also from a potential IR boundary. Because of the confinement property, the spacetime ends
smoothly in the IR, and the braneantibrane world-volume must end before or at the IR end of
space. The contribution to the gauge variation of the CS action coming from this potential IR
boundary would give rise to an additional contribution to the gauge anomaly, and make the bulk
theory inconsistent. It must therefore vanish. One is led then to the consistency condition
at the IR, which leaves us with the UV contribution alone:

iNc
iNc
=
Tr (L R )
SCS = ()p
(3.17)
5
(2)5 3
24 2
M5
Mink4
where we have used the fact that the tachyon vanishes on the UV boundary and reinserted a
(2 )3 factor in (3.15). We have defined the 4-form as

1
i
()p+1 F F + (A A F + A F A + F A A) A A A A .
2
2
(3.18)
Notice that the computation is rather similar to the one in [7], but here we have included the
dependence on the tachyon.9
In fact, we have checked the anomalous behavior under a transformation within the U (Nf )V
symmetry preserved by the vacuum. It is possible to find the anomaly in the full U (Nf )L
U (Nf )R chiral symmetry before it is spontaneously broken. For that purpose, we must consider
a more general ansatz in which T is proportional to a unitary matrix T T = T T = 2 INf . With
this milder assumption, the generalization of (3.12) reads:

Str ei F 6-form

1
= Tr eT T iFL FL FL + iFR FR FR
6
+ DT DT FL FL + DT FR DT FL DT DT FR FR
i
i
+ DT DT DT DT FL DT DT DT DT FR
4
4

1
DT DT DT DT DT DT .
(3.19)
60
One should now follow the same procedure as above and generalize the expressions (3.13)
(3.15). However, it proves a difficult task to find the five-form 5 . In any case, the IR term will
again vanish when due to the overall exponential factor. Since in the UV 0, one
can find the UV contribution even without knowing the full expression for 5 . The result is of
9 Notice the different definition of the gauge field with respect to [7]. A = i A,
F = i F where A,
F are the gauge field
and field strength used in [7].
113
course (3.17) but now with independent parameters for the left and right gauge transformations.

iNc
SCS =
(3.20)
Tr[L L R R ].
24 2
Mink4
Possibly up to a sign (which is a matter of conventions), (3.20) reproduces the QCD chiral anomaly, as we now check by using the general procedure sketched in [2] and further developed
in [40]. The AdS/CFT conjecture [1,2] states that Scl [A] = W [A] where W is the generating
W
functional for current (J A = A
A ) correlators in the boundary theory. We can equate the gauge
W
A
variation of both terms in the equality: Scl [A] is given in (3.20) while W [A] = A
A A =
4
4
A
A
A
A
d x D J = d x (D J ) . The anomalous divergences of the U (Nf )L,R flavor
currents in the presence of sources are, thus:
Nc
Tr(L,R ),
24 2

Nc
a
Tr a L,R .
D JL,R = ()hL,R
(3.21)
2
24
Here stands for the four-dimensional Hodge dual, we have defined hL and hR to be 0 and 1
respectively, and the currents have been decomposed in their Abelian and non-Abelian components following the conventions of Appendix A. Expressions (3.21) reproduce the known QCD
results (see for example [41]).
For completeness, we report here the explicit expressions of the traces appearing in
Eqs. (3.21). They can be easily derived from the definition (3.18) of :

1 ab a
p+1 1 a
a
U (1)
U (1)
b
c
Tr() = ()
(3.22)
F
f cA A F
F F + Nf F
2
8
U (1)
JL,R
and
= ()hL,R

1
Tr a = ()p+1 d abc F b F c + F a F U (1)
2

1 bc b
f a 3A Ac F U (1) 2AU (1) Ab F c
8

1 ab ecd
f e d f ac e d ebd + f bc e d ead Ab Ac F d
8

1 bc de ahl b
c
d
e
+ f hf l d A A A A .
16
(3.23)
3.3. The U (1)A axial symmetry

The first equation in (3.21) for the anomalous divergence of the Abelian flavor currents in the
presence of external sources is not complete. In the case of the U (1)A symmetry, an additional
anomaly appears [24].
To study this U (1)A anomaly, we need to know what bulk field couples to G G on the
boundary (here G denotes the field strength of the non-Abelian color gauge field). We consider
then a probe Dq color brane. Since we are studying four-dimensional gauge theories, the Dq
brane wraps a (q 3)-cycle Kq3 of the internal geometry. Typically, for a background dual to a
114
confining color gauge theory with no moduli space nor low-energy massive scalars, the Dq-brane
embedding is stable only at the boundary, in the far UV region.
The action for this probe brane reads:

d q+1 y e det P [g] + G + Tq
C Tr eiG ,
Sprobe = Tq
(3.24)
q+1
q+1
where P [g] is the pull-back of the background metric on the world-volume q+1 of the Dqbrane, and as before we have set 2 = 1. The coupling to G G we are interested in
T
comes from the expansion of the exponential in the WZ term, and is given by 2q q+1 Cq3
Tr(G G).
Comparison of this term to the usual YangMills action gives the holographic definition for
the QCD -angle (recall that p + q = 12)

C9p ,
QCD = 4 2 T12p
(3.25)
K9p
where the integral over K9p is evaluated in the far UV (the boundary).
To study effects on the dynamics of flavors related to Tr G G, we will need, therefore, to turn
on the RR potential C9p in the background, and look at its interaction with the flavor branes.
Using Hodge duality, this potential has the same degrees of freedom as Cp1 . The coupling of
this potential to the DpDp flavor branes can then be read from the WZ action (3.1)

Cp1 Str ei F 2-form .
SWZ = iTp
(3.26)
p+1
If we take T to be proportional to a unitary matrix, T T = 2 INf , as we did at the end of

Section 3.2, it is easy to write from (3.2) an explicit formula for Str ei F |2-form :

2
Str ei F 2-form = e Tr iFL iFR DT DT .
(3.27)
The coupling in (3.26) modifies the equation of motion of Cp1 , or equivalently the Bianchi
identity for the Hodge dual C9p . Indeed, the terms of the overall action involving Cp1 are
1

iF
S = 12
2 Fp Fp iTp p+1 Cp1 Str e |2-form , from which we obtain
2(10)

2
d F10p = d Fp = 2i(10)
Tp 9p (p+1 ) Str ei F 2-form
(3.28)
where F10p is the gauge-invariant (with respect to world-volume gauge transformations)

(10 p)-form RR field strength, and 9p (p+1 ) is the volume form for the space transverse to
the flavor branes times a -function localized at the position of the flavor branes.
From (3.28) we can write10 :
2
Tp 9p (p+1 ) 1
F10p = dC9p + 2i()p+1 (10)
(3.29)
10 The following formula has been computed for the case where all the quarks have the same (possibly vanishing) mass,
which is the situation considered in this paper. Since in Section 4.4 we will relate 1 to the would-be Goldstone boson
, we should expect that in a more general case Tr(log T log T ) will be substituted by an expression of the form
log(det T (T )1 ). We will not pursue this computation here.
115
where 1 is defined by d1 = Str ei F |2-form and can be explicitly obtained from (3.27):

2
1 = e Tr iAL iAR + log T log T d .
(3.30)
As we show in Appendix D, the only gauge transformations under which 1 transforms nontrivially are the U (1)A axial ones. We thus consider an infinitesimal such transformation L =
R = iINf :

2
2
1 = 2iNf e d d 2 = d 2iNf e = d0
(3.31)
where we defined 0 = 2iNf e .

Since F10p is gauge invariant, and as we just showed 1 = d0 , it follows from (3.29)
that the RR-potential cannot be invariant under flavor brane world-volume gauge transformations [39]:
2
2
C9p = 2i(10)
Tp 9p (p+1 )0 ,
(3.32)
2 = 1 (in units of 2 = 1),

which, inserted in (3.25) and using the identity 8 2 T12p Tp 10
results in the following transformation relation for the QCD theta angle under U (1)A transformations:
QCD = i0 |UV = 2Nf .

(3.33)
4 QCD
Because of the boundary coupling d x 8 2 Tr(G G), the formula (3.21) for the divergence
U (1)
U (1)
U (1)
JL
JR
has to be corrected in order to take into account
of the U (1)A current JA
U (1)
the non-trivial transformation of QCD (the U (1)V current JV
is unaffected):
U (1)
JA
Nf 1 ...4
Nc
Tr(L + R ) +

TrSU (Nc ) (G1 2 G3 4 ).
2
24
16 2
(3.34)
3.4. A holographic view of the ColemanWitten theorem

In [23], Coleman and Witten proved that, under a few assumptions, in Nc massless
QCD, chiral symmetry is spontaneously broken U (Nf )L U (Nf )R U (Nf )V .
One of the main results of the previous section is that, if the theory is confining, the tachyon
has to get a (diverging) vev near the IR even if T 0 in the UV. Since T transforms in the bifundamental representation of the flavor group, T = 0 means that the chiral symmetry is broken.
We showed in Section 2.2 that T is proportional to the identity matrix, and consequently the
flavor symmetry is broken down to U (Nf )V . Confinement, therefore, implies spontaneous chiral
symmetry breaking, reproducing in a holographic language the result of [23].
To make the analogy between these two different approaches more evident, we can compare
the assumptions of [23] with those we made in the present holographic setup. In order to do
this, we start by reviewing the five assumptions of [23]: (1) The large Nc limit of QCD exists.
(2) In that limit the theory is confining. (3) There is a single order parameter for the breaking
that is a quark bilinear and transforms in the bifundamental of U (Nf )L U (Nf )R . (4) Its vev
can be found by minimizing some potential. (5) The potential is assumed not to have degenerate
minima.
For the holographic setup, assumptions (1) and (2) are also necessary, and of course, one must
further assume the (non-trivial) fact of the existence of a holographic string dual of large Nc
116
QCD which fits in the general framework we have introduced. It is appealing that assumption (3)
is automatic in the setup of our work since it is a well established fact that the lowest state of a
braneantibrane system is the open string tachyon. Concerning assumption (4), the string theory
dual provides, in principle, a way to determine the vev of the quark bilinear: the bulk vev of the
tachyon field is determined by solving the bulk field equations with the UV boundary condition
corresponding to mq = 0 (2.9) and the IR consistency condition |IR . Then, one can read
the value of the quark bilinear vev from the UV behavior of (z) (2.9). This computation
requires the DBI part of the action, and is similar to that in [3,4]. It was discussed in Section 2.2.
In principle, one expects that the UV and IR conditions determine uniquely the bulk vev of the
tachyon and therefore the qq
condensate of the dual theory. Nevertheless, in the very general
setup we are considering, we cannot prove that the solution for the qq
is indeed unique, so this
must be taken as a further assumption, analogous to the non-degeneracy (5) listed above.
We remind the reader that in Section 2.2 we also made the (reasonable) assumption that the
stable minimum given by the vacuum (2.10) is the real vacuum of the theory.
The demonstration presented in this paper, namely that confinement is a sufficient condition
for spontaneous chiral symmetry breaking, can be thought of as a reformulation of the geometrical picture of [9]. There it was argued that if one places branes and antibranes in a background
that smoothly ends at some point of the radial coordinate (and therefore is confining), they must
necessarily recombine in the IR, sparking the breaking of chiral symmetry. The advantage of the
present formulation stands in the fact that the tachyon, the scalar responsible for the symmetry
breaking, is explicitly taken into account which, for instance, allows to introduce a bare quark
mass. Notice that, if one is in the conformal window of N = 1 or N = 0 QCD, then there is no
IR boundary and it is consistent to take a vanishing vev for the tachyon as in [19,25]. However,
as in [23] or [9], our argument does not rule out having spontaneous chiral symmetry breaking
without confinement.
3.5. The effects of a non-trivial quark mass
In [40], it was explained how an expression like (3.21) can be related to the anomalous threepoint function in the case of N = 4 SYM where the global symmetry associated to the current J
is SU (4)R . Analogously, in our case one can relate (3.21) to the anomalous massless QCD threepoint function. This anomaly equation in the large Nc limit was used by Coleman and Witten [23]
to prove the existence of massless Goldstone bosons, and therefore the spontaneous breakdown
of the symmetry U (Nf )L U (Nf )R U (Nf )V . Clearly, if there is a non-zero bare mass for
the quarks, there cannot be spontaneous breaking nor massless Goldstone bosons, so (3.21) must
be modified when mq = 0. In Section 4.3, we will show by analyzing the equations that define
the mesonic spectrum that there are massless Goldstones if and only if mq = 0. The goal of this
section is to describe holographically the modification of (3.21).
The axial current is classically non-conserved when mq = 0. In the holographic picture this
shows up because a gauge transformation of the bulk gauge fields is accompanied by a gauge
transformation of the tachyon (A.6), which contributes to (3.21) iff mq = 0. In the rest of this
section we make these statements more precise, assuming that the space is asymptotically AdS:
gxx (z)
2
RAdS
+ ,
z2
gzz (z)
2
RAdS
+ .
z2
(3.35)
The antihermitian part of the tachyon matrix T T

is dual to (i times) the pseudoscalar
2
current, see Table 1. We consider a perturbation around the vacuum T = ei where is pro
117
I + a a . At first order, T T
= i . There
2
4
must exist a boundary coupling contributing in the W generating functional d x (i0A J5A )z=0 ,
5 q + i q
5 a q and 0 is proportional to i .
where J5 is the pseudoscalar (pion) current J5 = i q
As argued in Section 2.2, we require that the mass of the tachyon, at least near the boundary, is
2 = 3 since qq
and i q
5 q are operators of dimension 3. Following Wittens prescripm2T RAdS
portional to the (hermitian) pion matrix,
2Nf
1
2
tion [2], one has to couple to the pseudoscalar current in the boundary, 0 = RAdS
z ican , since
for a scalar of the above cited mass, the non-normalizable behavior is z. Inserting the value of
can given in (2.9), the contribution to the generating functional is:

d 4 x mq A J5A .
(3.36)
We now compute how this expression transforms under an axial gauge transformation
L = R = = iA A = iI + ia a . From (A.6), the infinitesimal gauge transformation V e of the tachyon around the vacuum (2.10) is given by T = R T T L =
(R L ) = 2 , which in turn can be expressed as a transformation of the pion matrix

U (1)
2a J5a ).
as: A = (2i)A = 2A , which, inserted in (3.36), yields d 4 x mq (2J5
This leads to a modification of (3.21) and (3.34):
U (1)
JA
U (1)
= 2mq J5
Nc
Tr(L + R )
24 2
Nf 1 ...4

TrSU (Nc ) (G1 2 G3 4 ),
16 2

Nc
a
Tr a (L + R ) .
D JA = 2mq J5a +
(3.37)
2
24
One can now relate these expressions to the three-point function in the presence of a non-trivial
quark mass. The term with mq in (3.37) results in an extra term for the three-point function. The
presence of this new term invalidates the argument that Coleman and Witten used in the mq = 0
case to show that the three-point function has a pole at zero momentum. Thus, as expected, the
three-point function can be analytic at zero momentum, consistent with the absence of massless
Goldstone bosons.
+
4. Features of the mesonic mass spectrum

4.1. Equations determining the spectrum
We now study small fluctuations of the different fields around the vacuum configuration. They
correspond to the mesons of the field theory. We will just consider quadratic terms in the action,
which are enough to find the mass spectrum, but not the couplings. At this level, the non-Abelian
action (2.4) is just the sum of Nf2 copies of the Abelian one (2.1). Thus, we use (2.1) in the
following and there are Nf2 copies of each of the towers of resonances that we will study.
We start by considering excitations for the fields , AL , AR around the vacuum of Section 2.2.
Fluctuations of the modulus of the tachyon will be briefly mentioned below. We expand the
I and assuming there is no B-field). We
simplified version of (2.1) (neglecting the scalars Y(i)
define:

2
e = eeff V 2 ,
(4.1)
g zz = gzz + (z )2 .
118
In order to use a notation similar to the one in [13], we choose a gauge:

R
AL
z = Az = 0
(4.2)
and define:
VM
AL + AR
M
= M
,
2
AM
AL AR
M
= M
,
2
2 2
v = ,
and V , A as the (Abelian) field strengths of V , A . Expanding (2.1), we find:

1
1 12
S = d 4 x dz e g zz
V V + A A + gxx g zz 2 (z V )2 + (z A )2
2

1
1
1 2 12 2
2
2
2 2
+ gxx g zz v ( + 2A ) + gxx g zz v (z ) .
4
4
(4.3)
(4.4)
Vector sector
The vector (1 ) sector is decoupled from the rest in the action (4.4). The equation of motion
for V can be solved by expanding V in modes:
V =

(n) (z)V(n) x
(4.5)
with:

1
21
z e gxx g zz 2 z (n) + m2n e g zz
(n) = 0.
(4.6)
The physical modes are those which yield a finite action when integrating in z. Substituting (4.5)
in (4.4), one finds a tower of massive vectors:

1
(n)
S = d 4x
(4.7)
V(n) + m2n V(n) V (n)
V
2
n
provided the normalization conditions11 :
zIR
1 21 2
1
dz e g zz
(n) = ,
2
2
0
zIR
dz e gxx g zz 2 (z (n) )2 = m2n .
(4.8)
Integrating by parts the second expression and using (4.6), we consistently obtain the first line
of (4.8).
11 We have chosen this normalization such that when the fields are promoted to their non-Abelian generalization V (n)
(n)a
(n)a
V a , the components V are normalized in the standard way, since we are using Tr a b = 12 ab . We apply this
prescription to all the different towers of modes.
119
Axial vector sector

The vector field fluctuation A can be split in a transverse and a longitudinal part, A = A
+

++ excitations. We
A , with A
= 0. We first consider the transverse part, corresponding to 1
expand it as:

(n)
A
A
(4.9)
=
(n) (z)B x
n
which results in a tower of massive axial vectors:

1
(n)
S = d 4x
B(n) + m2n B(n) B (n)
B
2
n
(4.10)
subject to the normalization conditions:

zIR
1 21 2 1
dz e g zz
A(n) = ,
2
2
0
zIR
1
2

1
2
2 2 2
dz e gxx g zz 2 z A
(n) + e gxx g zz v A(n) = mn
(4.11)
and the second order differential equation:

1
1

1
2 2
2 2
z e gxx g zz 2 z A
(n) + mn e g zz A(n) e gxx g zz v A(n) = 0.
(4.12)
It is interesting to notice at this point a difference with respect to constructions like the Sakai
Sugimoto model [7]. In that kind of models, the vector and axial vector mesons are solutions
of a single second order differential equation with different matching conditions at the IR. In
this case, they satisfy different equations (4.6), (4.12). If the chiral symmetry were not broken
(v = 0), the equations would become degenerate and vector and axial vector mesons would have
the same mass spectrum due to the unbroken symmetry. The differential equations found in this
construction coincide with those of the AdS/QCD models [13,14] up to the z-dependent dilaton,
metric factors and tachyon vev, which we have kept general.
Pseudoscalar sector
The modes coming from and the longitudinal part of A combine to give a single tower of
resonances with the quantum numbers of pions 0+ . The equations of motion from (4.4) can be
solved by expanding:

(n) (z) (n) x ,
A =
=2

(n) (z) (n) x
(4.13)
where the functions (n) and (n) satisfy the following coupled differential equations
1

1
2 2
v ((n) (n) ) = 0,
z e gxx g zz 2 z (n) + e gxx g zz
gxx v 2 z (n) m2n z (n) = 0.
(4.14)
120
Inserting (4.13) in (4.4) and normalizing:

zIR
1

1
2 2
dz e gxx g zz 2 (z (n) )2 + e gxx g zz
v ((n) (n) )2 = 1,
(4.15)
zIR

2 12 2
dz e gxx
g zz v (z (n) )2 = m2n
(4.16)
one finds a tower of pseudoscalars in the four-dimensional theory:

2

2
(n) + m2n (n) .
S = d 4x
(4.17)
Scalar sector
The scalar mesons (0++ ) come from excitations around the vacuum of the form = S(x , z).
Expanding (2.6) up to quadratic order in S, one finds an action:

2

1
1
2 V
1 V
2
2 2
2 2
g
g
S
+
gxx
g zz (z v)S(z S)
S = 2 d 4 x dz e
xx zz

2 V =v/2
2 V =v/2

1
1 2 32
1
1
2
+ gxx
(4.18)
g zz g zz (z v)2 (z S)2 + gxx g zz
( S)2
4
2
from which one can straightforwardly extract the linear equation of motion and normalizability
condition. Since we will not need them in the following, we do not present them explicitly.
4.2. On linear confinement (for highly excited mesons)
In a theory with linear confinement such as QCD, one expects that the squared masses m2n
of highly excited hadrons grow linearly with the excitation number n: m2n n. This behavior is
difficult to find from holographic models, that typically yield m2n n2 [42]. In [16] it was shown
that, by appropriately tuning the behavior of the metric and dilaton in the IR, one can indeed find
the expected behavior. Moreover, it was conjectured that such behavior could come from closed
string tachyon condensation, but no concrete model producing such IR asymptotics of metric and
dilaton has been found.
We will show here that the relation m2n n is automatic in our construction. Therefore, the
linear confinement relation for mesons comes from open string tachyon condensation on the
flavor brane world-volume. In this section we just state the ideas and results while details are
relegated to Appendix E.
The argument relies on the fact that the meson excitations feel the effective open string dilaton
and metric (see Eq. (4.1)) rather than just the closed string dilaton and metric. In the IR,
so g zz = gzz + 2 (z )2 2 (z )2 even if gzz has a single pole. Moreover, gxx goes to a constant,
which, on general grounds, can be
identified with the QCD string tension. Reinserting and
defining a new radial variable u = T2QCD (z), we have, near the IR, where u, :
ds = 2
2

2
TQCD dx1,3

2

2
2
2
+ (z ) dz = 2 TQCD dx1,3
+ du2 ,
121
TQCD 2

TQCD 2
2
e V 2 e e 2 u
(4.19)
u .
2
It turns out that the natural radial variable in the IR is proportional to . The quadratic growth
of the dilaton (4.19) is the IR behavior advocated in [16] to account for linear confinement. In
fact, using the WKB approximation to compute the masses of highly excited bound states one
obtains, for the vector meson tower (see Appendix E):
m2n 2TQCD n (large n, vector mesons).
(4.20)
This is a quite general result that only relies on having a confining background and therefore
diverging tachyon, on the large Gaussian behavior of the tachyon potential and on the DBI
action. We find very encouraging that our construction naturally accounts for the relation of
linear confinement, including the expected multiplicative factor.
However, an important remark is the following: if one repeats the same computation for axial
vector mesons, Eq. (4.12), one finds:

16
2
mn = 2 1 + 2 TQCD n (large n, axial vectors).
(4.21)
This disagrees with some expectations [22], since vector and axial mesons should asymptotically
have equal masses at large n if chiral symmetry is restored for excited hadrons. A potential caveat
of our approach is the fact that we are using the DBI action outside its range of validity. This
point definitely deserves a better understanding.
Finally, it is important to stress that in view of the confining IR behavior, there is no need to
impose an arbitrary IR boundary condition. The IR condition for physical excitations is simply
the normalizability of the action [16].
4.3. Goldstone bosons
Large Nc QCD with massless quarks has a set of Nf2 massless pseudoscalars which are the
Goldstone bosons of the spontaneous breaking of the U (Nf )A . We will generically call them
pions. In this section we show how they appear in this construction.12 They are the solutions to
Eqs. (4.14)(4.16) with mn = 0. Notice that, as expected, there cannot be massless (axial) vectors
due to (4.8), (4.11).
We generalize the analysis in (4.13)(4.17) with mn = 0 to the non-Abelian Nf > 1 case.
Define the pion matrix as the generalization of the (0) (x ) in (4.13):

(x )
(x ) =
I + a x a .
2Nf
To lowest order, the action for the pions is just the one for a set of massless scalars:

S = d 4 x Tr

1
1
= d 4 x + a a .
2
2
12 One of the N 2 pseudoscalars is massless only in the strict N limit. This is described in Section 4.4.
c
f
(4.22)
(4.23)
122
Due to (4.14), (0) must be a constant. We will shortly see that it is related to the pion decay
constant. We define a quantity:
(z) = (0) (z) (0)
(4.24)
which, regarding (4.14), (4.15), satisfies:

1

1
2 2
z e gxx g zz 2 z (z) e gxx g zz
v (z) = 0,
1

e gxx g zz 2 (z)z (z)z=0 = 1.
(4.25)
(4.26)
For the second line, we have integrated by parts (4.15) and taken into account that for (z) we
must choose the normalizable mode in the IR.
The question of whether there are massless modes boils down to the existence of a function
(z) satisfying (4.25), (4.26). From (4.26), it is obvious that the answer depends
on the UV
2 2
behavior of (z), which, due to (4.25) depends on the UV behavior of v = .
We now give a heuristic argument suggesting that if quarks are massless, there are indeed
massless Goldstone bosons. Since is dual to qq,
the UV behavior of is related to the quark
mass and condensate. If there is no quark mass, is normalizable, i.e. it vanishes fast enough in
the UV. It is natural to think then that the second term in (4.25) can be neglected near the UV,
allowing for a solution to (4.25) and (4.26), which is, asymptotically,
1
f (z).
(z) = (0) + (0)
(4.27)
The constant (0) has to be determined by the IR condition, we have defined f (z) such that
1/2
z f (z) = (e gxx g zz )1 and used the residual gauge invariance to fix (0) (0) = 0 and therefore f (0) = 0. If the quarks are, instead, massive, v is larger in the UV and the second term in
(4.25) cannot be neglected, which hinders the existence of a solution as the one just described
corresponding to a massless Goldstone.
Just as a clarifying example, we consider a five-dimensional space which is asymptotically
AdS. In this case, all expressions asymptote in the UV to those of [13], where one can explicitly
check that the heuristic reasoning above is valid.
The pion decay constant
Consider mq = 0 so that there are indeed massless pions. The pion decay constant can be
related to the pole of the axial current two-point function at zero momentum:

A q 2 =
n
fA2n
q 2 + MA2 n
f2
.
q2
(4.28)
One can compute this correlator following the AdS/CFT prescription, as in [13,14]. This is done
by computing the on-shell action giving the appropriate boundary condition to the field dual to
2
the axial current, i.e. A

. When q = 0, the bulk equation for A (z) is (4.12):
1

1
2 2
z e gxx g zz 2 z A
(n) e gxx g zz v A(n) = 0.
(4.29)
(n)
Substituting in (4.4) and deriving twice with respect to B (as defined in (4.9)), one gets:

1
f2 = lim e gxx g zz 2 A (z)z A (z)|z=

0
(4.30)
123
where A (z) is a solution to Eq. (4.29) subject to the UV condition A |z= = 1 and the IR
normalizability condition. Notice there is an extra factor of 1/2 in (4.30) with respect to what
one would get from (4.4). This is again because we use conventions suitable for the non-Abelian
generalizations of the fields, see footnote 11.
Two remarks are in order: even if the f in (4.30) depends on quantities at z = 0, the value of
f depends on the full metric, since one has to select the well behaved IR mode. This typically
involves a non-trivial numerical integration. Notice also that f2 is of order Nc since inside e
there is a D-brane tension which scales as gs1 Nc .
Comparing (4.25), (4.26), (4.29), (4.30) and using the fact that both (z) and A (z) must
follow the normalizable behavior in the IR, it is straightforward to conclude that up to an unimportant sign, (z) = f1 A (z). This equality at z = 0 yields:
(0) =
1
.
f
(4.31)
The Gell-MannOakesRenner relation

We now show how, considering asymptotically AdS, one can obtain the GOR relation [43],
which gives the masses of the pions when the quark mass is small but non-vanishing. The argument is very similar to the one in [13] but details are different. In the following we assume
mq = 0 but only keep terms linear in mq .
We have to solve (4.14)(4.16) for small mn m . This solution can be obtained as a perturbation of the mn = 0 case studied at the beginning of this section. As opposed to that case,
cannot be a constant and, from (4.16), one can show that |z=0 = 0 is needed. Then, using (4.14):
z
(z) = m2
0
z (z) m2 f
dz
=
2
gxx v 2
z
dz
0
z3
.
(mq z + z3 )2
(4.32)
For the second equation we have used that the integral is dominated by the small z region so we
can substitute the asymptotic value of the different functions. To obtain z (z) we have used its
value in (4.27) and for v 2 we have substituted (4.3) and (2.9). As a consistency
check, notice that
the region where the integral above has significant support is around z mq / , so taking mq
small enough,
the integral only probes the asymptotically AdS region.
For z mq / , the function (z) goes to a constant which has to be the one of the massless
case (4.31). Putting everything together, we get the known expression:
m2 =
4mq
2mq qq
=
2
f
f2
(mq 0)
(4.33)
where we have substituted (B.10).

4.4. The mass of the
We now return to the mq = 0 case, where, in principle there are Nf2 massless Goldstones. In
N
Section 3.3 it was shown how our setup correctly reproduces the O Nfc anomaly of the U (1)A
axial symmetry [24]. This anomaly implies that the (generalization of the) meson, the wouldbe Goldstone boson corresponding to the diagonal U (1)A subgroup of the spontaneously broken
N
U (Nf )A , has a mass of order Nfc , and is therefore massless only in the strict Nc limit.
124
The mass appears in the present holographic setup via a Stuckelberg mechanism. The reasoning we will follow is very similar to the one of [7] (see [6,44] for related work and [45] for a
recent discussion in deconfined theories with broken chiral symmetry).
The C9p can only appear in the action in the gauge invariant combination (3.29). Integrating
this expression in the space composed of the 9 p non-Minkowski dimensions that the color
branes wrap plus the radial direction we obtain:

(2)2 T12p
F10p = QCD + i()p+1 (1 )z dz
M10p
2Nf
.
f
= QCD + ()
(4.34)

We have used (3.25) and M10p dC9p = K9p C9p where the last integral is evaluated in the
UV. For a 3 + 1 confining theory defined on Dq-branes with q 4 as in [34,46] this equality
holds automatically because the space closes off smoothly at the IR and K9p is the boundary
of M10p . On the other hand, if one wants to build some kind of five-dimensional model in the
spirit of [1317], one would have to impose that the RR-potential vanishes in the IR boundary to
find the same condition.
i 2(x )
The integral in (4.34) is computed substituting T = ei = ei2(x ) = e f in (3.30):

2
(1 )z dz = e Tr (log T log T ) d
4i
=
Tr()
f

e
d = i
2Nf
.
f
(4.35)
The contribution of QCD to the vacuum energy density can be computed by integrating the
kinetic term of the RR C9p -form [47], and must appear through the gauge invariant combination
(4.34). Thus:

2
4
p 2Nf
S=
(4.36)
d x QCD + ()
2
f
where the topological susceptibility of the vacuum is, by definition, the second derivative of
2
the vacuum energy density in the glue theory without flavors = d 2 E no quarks . It can be
dQCD
computed from the supergravity action [47]. The expression (4.36) reproduces the Veneziano
Witten formula for the mass of the [48]13 :
m2 =
2Nf
.
f2
(4.37)
5. Comments and discussion

Compared to the models [3] ([4]) where flavors where introduced with D7 (D6)-branes in the
background of D3 (D4)-branes, our construction has the advantage that the breaking of the full
13 The difference in a factor of 2 with respect to [48] is due to a factor 2 in the definition of f . In our conventions

0|J U (1) | = 2Nf p f and f = f up to suppressed O(Nc1 ) corrections.
125
non-Abelian chiral symmetry is described. In fact, one may think of the models presented in
those papers as the result of a DpDp system in which the tachyon has fully condensed leaving
a D(p 2) brane as a vortex. Since a non-trivial tachyon breaks the U (Nf )A , this is not present
in the final configuration and only a remnant U (1)A is left as the rotation symmetry around the
vortex (for a related discussion, see [18]).
The model of Sakai and Sugimoto [7,8] and its generalizations share many properties with
our setup, but there are also important differences. In [7,8], it does not seem possible to include
a bare mass for the quark, so clearly there are aspects of the QCD symmetry breaking that are
not present in that setup. Another phenomenon that we have shown to be intimately related to
the physics of the tachyon is the existence of massive resonances with the quantum numbers of
pions 0+ . These are not present in the model of [7,8]. A heuristic way of understanding the
differences is that the description of [7,8] is done already in the broken phase (with the brane
antibrane pair reconnected), so that information about the order parameter is lost. It would be
interesting to try to include the tachyon in such type of models, generalizing the formalism of
this paper to non-overlapping braneantibrane with excited transverse scalars. For some progress
in this direction, see [11].
In a sense, the model presented in this paper can be thought of as a string theory construction
in which the AdS/QCD models [13,14] are embedded. We have seen that the five-dimensional
field spectrum is the same. Expanding the square roots for small values of AM and T , one gets
the same kind of action as in [13,14]. In fact, the equations to determine the spectrum (4.6),
(4.12), (4.14) reduce to those in [13,14] if one considers AdS metric and constant dilaton. On the
other hand, the equation for the scalar mesons (4.18) is different from its analog in [49]. Small T ,
as argued above, corresponds to the UV. The successful features of [13,14] come from the UV.
We therefore conclude that a model built along the lines described in this paper in asymptotically
AdS space can capture the good physical features of [13,14] while the IR arbitrariness of those
models is lifted due to the condition |IR . This condition fixes the quark condensate in
terms of the quark mass and removes the extra non-physical parameter appearing in [13,14].
Moreover, the brane physics automatically provides the 5d ChernSimons term.
By applying the formalism described in this paper to a concrete model (i.e. to particular expressions of the metric and dilaton), one could find numerical estimates of QCD observables such
as meson masses or couplings. Nevertheless, as stressed several times, enforcing a well behaved
2 = 3 because the quark bilinear has dimension 3.
asymptotically AdS space, implies m2T RAdS
1
2
2

Since mT = 2 , we need RAdS = 6 and the space has large curvature, of the order of the
string scale. The complete dual background, then, is not a solution to just Einstein equations,
but higher derivative corrections to the equations of motion have to be introduced. This is a very
difficult, if at all possible, task.
The drawback of large curvature implies that meson masses or couplings which would be
numerical results obtained from the DBI action cannot be considered trustworthy. This same
problem has arisen when trying to build non-critical holographic models. Nevertheless, Einsteinlike equations have been used to extract qualitatively correct results [10,25,50]. Considering the
impressive quantitative precision of predictions in models like [7,8,13,14], it would be interesting
to build at least phenomenological models incorporating the tachyon physics. Constructions like
[15,17] could be a starting point.
In any case, apart from quantitative computations, we expect the general features derived from
the DBI action to hold. Moreover, since the WZ term is topological, the results derived from its
analysis in Section 3 hold even if the curvature is large.
126
We end by commenting on some additional open problems. First of all, it would be of major
interest to add finite temperature and describe the deconfinement phase transition. The physics of
fundamental matter in this regime is very rich and has been studied holographically (see [51] and
references therein). It may be possible to generalize the analysis to the kind of setup described in
this paper. It would also be nice to understand the physics of rotating strings and Wilson loops.
Another possible generalization is to introduce a large number of flavors Nf Nc , to go beyond
the quenched approximation. This may be done along the lines of [19,25,52]. In fact, the general
form of our expressions should guarantee that they can account for any such backreaction, as
long as the theory with fundamentals is still confining. Finally, it would also be interesting to
make contact with holographic N = 1 theories built on the cigar, where some exact world-sheet
computations can be done despite the large curvature, [26].
Acknowledgements
It is a pleasure to thank D. Aren, M. Berg, F. Bigazzi, A. Cotrone, U. Gursoy, D. Mateos, V. Niarchos, F. Nitti, A. Pomarol, A.V. Ramallo and J. Sonnenschein. R.C. and A.P. are
supported by European Commission Marie Curie Postdoctoral Fellowships, under contracts
MEIF-CT-2005-024710 and MEIF-CT-2005-023373, respectively. This work was also partially
supported by INTAS grant, 03-51-6346, RTN contracts MRTN-CT-2004-005104 and MRTNCT-2004-503369, CNRS PICS # 2530 and 3059, and by a European Union Excellence Grant,
MEXT-CT-2003-509661.
Appendix A. Conventions
For the U (Nf ) generators A , A = 0, . . . , Nf2 1, we take
A
= A ,
1

Tr A B = AB .
2
(A.1)
This in particular fixes the normalization of the U (1) generator: 0 = 1
2Nf
Nf Nf identity matrix. For the non-Abelian SU(Nf ) generators

have
a b

, = if ab c c ,
Tr a b , c = d abc
a ,
I, where I is the
a = 1, . . . , Nf2 1, we
(A.2)
where f ab c and d abc are, respectively, the structure constants and the normalized anomaly
Casimir for SU(Nf ). Because of (A.1) f ab c and d abc are real numbers.
We define gauge fields to be hermitian
U (1)
I + Aa a .
A = AA A = A
(A.3)
In differential form notation, the field strength and covariant derivative read then
F = dA iA A,
D d iA,
(A.4)
where A indicates the representation-dependent action of the gauge algebra. In particular the
Bianchi identity reads DF = 0, and the covariant derivative of the tachyon is given by
DT = dT + iT AL iAR T ,
DT = dT iAL T + iT AR .
(A.5)
127
Under gauge transformations, the left and right gauge potentials, and tachyon transform in the
following way
AL VL AL VL idVL VL ,
T VR T VL ,
AR VR AR VR idVR VR ,
T VL T VR ,
VL VL = VR VR = I.
(A.6)
(x) = e(x)
An infinitesimal gauge transformation is defined as V

1 + (x) and the gauge
transformation of a field as A A + A. From (A.6) we have then:
A = iD = i d + [, A],
F = [, F ],
L T = T L ,
R T = R T .
(A.7)
Notice that the generators of gauge transformations are antihermitian. When we decompose them
in their U (1) and SU(Nf ) parts, we will write then
= iA A = iI + ia a
(A.8)
with and a real parameters. In particular we have, from (A.3) and (A.8)
U (1)
A
=
and Aa = (D )a .
(A.9)
Currents are defined to be hermitian. We decompose a U (Nf ) flavor current as

J = J0 0 + Ja a .
This decomposition corresponds to defining the Ath component as

A
1 5 A
JL,R = Trcolors i q
q .
2
For the U (1) component J 0 , this would be a strange normalization

1 5
1
0
JL,R
=
Tr i q

q .
2
2Nf

We therefore define a rescaled J U (1) = 2Nf J 0 , such that J now reads
J =
1 U (1)
J
I + Ja a .
2Nf
(A.10)
(A.11)
(A.12)
(A.13)
Notice that the normalization of AU (1) in (A.3) has been chosen in such a way that the boundary
coupling of the current to the gauge field reads

4
U (1)
+ J a Aa .
2 d x Tr J A = d 4 x J U (1) A
(A.14)
Appendix B. Determination of qq
from holographic renormalization
When solving the equation for the modulus of the tachyon in asymptotically AdS space, we
found that it depends on two integration constants mq and , see (2.9). The constant associated to
the non-normalizable mode mq can be immediately identified with the quark mass. On the other
hand, is related to the quark condensate qq
in, in principle, a non-trivial way. Schematically,
S
qq
= m
,
where
S
denotes
the
on-shell
action. However, the on-shell action is UV divergent,
q
128
and must be renormalized by adding covariant counterterms. This is done by following the socalled holographic renormalization procedure (for a review, see [53]). In the following, we adapt
to our case the method of [54], where holographic renormalization was applied to probe flavor
branes. In fact, it will be enough for our purposes to consider a simple case in which the scalar
does not depend on the Minkowski x -coordinates. Also for simplicity, we consider that the
metric and dilaton only depart from their AdS values at an order that does not contribute to
2 /z2 (1 + O(z5 )), g (z) = R 2 /z2 (1 + O(z5 )) and eeff =
UV divergences, i.e. gxx (z) = RAdS
zz
AdS
e0 (1 + O(z5 )).
We write the action in terms of a canonically normalized, rescaled tachyon, as defined in (2.9):
= ccan ,
c2
e 0
.
4Tp
The action (with a UV cutoff ) in terms of these quantities reads:

zIR
5
RAdS
1 c2 can
2
2
4
d x dz 5 e
1 + c2 z2 (z can )2 .
Sreg =
2
3
2c
z
(B.1)
(B.2)
In the asymptotically AdS region, the equation of motion for can reads:

c2 (z can )z (z2 (z can )2 )
=0
3can z5 z z3 z can + z2
3
1 + 23 c2 z2 (z can )2
and can be solved in series for small z as:

4
1 2 2
3
2
can = z (0) + c z log z(0) + z (2) + O z .
3
(B.3)
(B.4)
Inserting back this result in the on-shell action, one finds UV divergencies as 0. They have
to be subtracted by adding the following counterterms localized at a z = slice:

RAdS
RAdS
4
2
d 4 x can
d
x
,
S
=
,
Sct0 =
ct1
2
6
8c
RAdS c2
4
d 4 x (log )can
Sct2 =
(B.5)
,
18
4 4 . As
where is the determinant of the induced metric in the z = slice, i.e. = RAdS
pointed out in [54], one can also add a finite counterterm:
4
Sct3 = d 4 x can
(B.6)
where is some constant. It was argued in [54] that different values of correspond to different
renormalization schemes. Defining:
Ssub = Sreg + Sct0 + Sct1 + Sct2 + Sct3
(B.7)
the quark condensate is given by:

Ssub
32
qq
= lim RAdS
0
can ()

2
7
5
c RAdS
2
3
2
2
4 .
= RAdS (2) + RAdS (0)
3
3
(B.8)

3/2
129
3/2
2 = 6 =
Noticing from (2.9), (B.4) that (0) = RAdS mq ; (2) = RAdS and substituting RAdS
3
, we find:

3 2
qq
= 2 + mq c
4 .
3
3
(B.9)
In Section 4.3 we have used the value of the condensate for small mq which is unambiguously
given by:
qq
2
(mq 0).
(B.10)
Appendix C. Comments on the superconnection formalism

In Section 3 we have worked with the WZ world-volume action for stacks of braneantibrane
pairs and used a superconnection formalism which makes the notation quite compact. In this
appendix we review some definitions and properties
this construction. The supermatrices are
of
B
where the blocks A, B, C, D are Nf
2Nf 2Nf matrices of differential forms M = AC D
Nf matrices. We deal with two kinds of supermatrices: if the blocks in the diagonal A, D are
composed of odd (even) differential forms, then, the off-diagonal ones B, C consist of even (odd)
forms. For instance, A (F ) defined in (3.2) are matrices of each kind.
The multiplication of supermatrices is defined as [37]:

A B
A B
AA + ()c BC AB + ()d BD

=
M M =
(C.1)
C D
C D
DC + ()a CA DD + ()b CB
where x is 0 if X is a matrix of even forms or 1 if X is a matrix of odd forms. The associative
property of this product can be easily checked.
The supertrace is defined as

1 0
A B
Str M = Tr
M Str
= Tr A Tr D.
(C.2)
0 1
C D
It is straightforward to prove the cyclic property of the supertrace, where the A-type (F -type)
supermatrices behave as odd (even) forms:
Str(ab) = Str(ba);
Str(af) = Str(fa);
Str(fg) = Str(gf).
We have denoted by a, b (f, g) generic supermatrices of the A-type (F -type).

We also define a pseudotransposition operation:

pt t
A
iC t
A B
iB t D t
C D
(C.3)
(C.4)
where t denotes usual matrix transposed. This generalizes the usual transposition in the sense
that (using the same notation of (C.3)):
(ab)pt = bpt apt ;
(af)pt = fpt apt ;
(fg)pt = gpt fpt .
(C.5)
Nevertheless, unlike the usual transposition, pseudotransposing twice does not yield the initial
matrix. Obviously, the supertrace does not change under pseudotransposition.
130
Appendix D. Gauge transformation of the 1-form 1

In this appendix we show that the 1-form 1 defined in (3.30) in Section 3.3 transforms nontrivially only under U (1)A . This is an important point in the understanding of the U (1)A axial
anomaly and it is therefore worth being explicitly shown. For clarity of exposition, we report
here the expression (3.30) for 1

2
1 = e Tr iAL iAR + (log T log T ) d .
(D.1)
It is obvious that the first two terms transform only under U (1)A transformations. The remaining
log terms, instead, require a little more effort. First of all notice that we can write a general
variation of Tr(log T log T ) as

1

Tr log T log T = Tr T 1 T T T

1
2
= 2 Tr T T T T = 2 Tr T T ,
(D.2)
where in the second equality we used the same condition on T we imposed in Section 3.3,
T T = T T = 2 INf , from where it also follows that T T = T T , which we used in the
last equality of (D.2).
From (A.7), we may show that for a vectorial (L = R = V ) gauge transformation
V Tr(log T log T ) vanishes. For an axial infinitesimal transformation (L = R = A ),
from (A.7) and (D.2), we find, instead:

A Tr log T log T = 4 Tr A .
(D.3)
This is clearly non-zero for Abelian U (1)A axial gauge transformations only, as we wanted to
show.
Appendix E. On the WKB approximation and linear confinement
In order to use the WKB approximation to determine the spectrum, we start by mapping the
standard eigenvalue problem to a Schrdinger-like equation. We follow [42]. We start from an
equation (notice that n is always proportional to the square of the four-dimensional mass):

(z)
1
z (z) + B(z)(z) = n (z)
z
(E.1)
(z)
2 (z)
and define a new radial variable u and a rescaled wave function :

(z(u))
du
= (z),
(u) = (u) z(u) ,
(u)
.
dz
(z(u))
(E.2)
Then, Eq. (E.1) is rewritten as14 :

(u) + V (u)(u) = n (u)
(E.3)
14 In this appendix, we define V as the Schrdinger like potential that appears in Eq. (E.3). It should not be confused
with the tachyon potential used in the rest of the paper.
131
with:

(u)
+ B z(u) .
(u)
One can use the WKB approximation to estimate the mass of high excitations:
1
u2
du
dn
= 2
dn
n V (u)
V (u) =
(E.4)
(E.5)
u1
where u1 and u2 are the classical turning points. In [42] it was observed that, quite generically,
the range in the variable u is finite. Therefore, for large n , the problem is similar to an infinite
1/2
potential well: the integral behaves as n
and n m2n n2 . In [16] it was remarked that the
behavior of the functions defined above can be tuned near the infrared such that the IR turning
point in Eq. (E.5) goes to infinity in such a way that the desired relation m2n n is obtained.
We may naively expect that highly excited masses should depend mainly on the UV behavior of
different functions. However, linear confinement emerges from the IR dynamics and the size of
highly excited hadrons grows with the excitation number. Therefore, highly excited mesons are
still affected by IR physics.
We now apply this formalism to the vector meson
equation (4.6). Comparing to (E.1), we
1/2
zz
define n m2n , B(z) 0, (z) e g zz , (z) ggxx
. The change of variable to u would
involve a complicated integral, but in order to estimate the large excitation spectrum, it is enough
to study the leading UV and IR behavior of the functions.
2 /z2 , e const,
Near the UV, we consider an asymptotically AdS space g zz gxx RAdS
3
so 1 and uUV = z. The potential (E.4) is V UV (u) 4u2 . Thus, for large n , the classical
turning point is at u1 = 3/(4n ) so u1 remains finite as n and, as expected, the UV

1/2
contribution to the integral in (E.5) decreases as n .
The intermediate region when one cannot apply the IR nor the UV asymptotics has finite size
1/2
in the u-variable so its contribution to the integral also decreases as n .
Near the IR, the vev of the open string tachyon is diverging and gxx TQCD , g zz 2 (z )2
and
e e . By substituting in (E.2), we find that the u-coordinate near the IR is just uIR
2
2
TQCD (z) and the potential is VIR ( 2 TQCD u) . The classical turning point is therefore at
2
1/2
u2 = T2QCD n . Thus, u2 grows to infinity as n , unlike the cases considered in [42]. The IR
contribution to the integral is:
u2
du
n ( 2 TQCD u)2
1
+ ,
TQCD
(E.6)
where the dots stand for a piece that vanishes as n . Therefore, the integral in (E.5) is
dominated by the IR region and:
dn
= 2TQCD
n dn
lim
(E.7)
which, reinserting n = m2n yields (4.20) as we wanted to show. For the axial vectors, (see (4.12)),
the z-dependent mass for A (such that B(z) = gxx v(z)2 ) adds an extra term to the potential near
2 u2 , leading to (4.21).
the IR VIR = ( 2 /4 + 4)TQCD
132
Fig. 2. Qualitative behavior of the Schrdinger-like potential in the u-variable. Near the UV, it grows as V u2 while
near the IR V u2 . In the middle, it may present more complicated features which do not affect the leading behavior of
the spectrum for large excitation number.
Fig. 2 depicts a sketch of the Schrdinger-like potential in the u-variable.

References
[1] J.M. Maldacena, The large N limit of superconformal field theories and supergravity, Adv. Theor. Math. Phys. 2
(1998) 231, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200;
S.S. Gubser, I.R. Klebanov, A.M. Polyakov, Gauge theory correlators from non-critical string theory, Phys. Lett.
B 428 (1998) 105, hep-th/9802109.
[2] E. Witten, Anti-de Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150.
[3] J. Babington, J. Erdmenger, N.J. Evans, Z. Guralnik, I. Kirsch, Chiral symmetry breaking and pions in nonsupersymmetric gauge/gravity duals, Phys. Rev. D 69 (2004) 066007, hep-th/0306018.
[4] M. Kruczenski, D. Mateos, R.C. Myers, D.J. Winters, Towards a holographic dual of large-N (c) QCD, JHEP 0405
(2004) 041, hep-th/0311270.
[5] N.J. Evans, J.P. Shock, Chiral dynamics from AdS space, Phys. Rev. D 70 (2004) 046002, hep-th/0403279;
K. Ghoroku, M. Yahiro, Chiral symmetry breaking driven by dilaton, Phys. Lett. B 604 (2004) 235, hep-th/0408040;
N. Evans, J. Shock, T. Waterson, D7 brane embeddings and chiral symmetry breaking, JHEP 0503 (2005) 005,
hep-th/0502091;
R. Apreda, J. Erdmenger, N. Evans, Scalar effective potential for D7 brane probes which break chiral symmetry,
JHEP 0605 (2006) 011, hep-th/0509219.
[6] J.L.F. Barbon, C. Hoyos, D. Mateos, R.C. Myers, The holographic life of the eta, JHEP 0410 (2004) 029, hepth/0404260.
[7] T. Sakai, S. Sugimoto, Low energy hadron physics in holographic QCD, Prog. Theor. Phys. 113 (2005) 843, hepth/0412141.
[8] T. Sakai, S. Sugimoto, More on a holographic dual of QCD, Prog. Theor. Phys. 114 (2006) 1083, hep-th/0507073.
[9] O. Aharony, J. Sonnenschein, S. Yankielowicz, A holographic model of deconfinement and chiral symmetry restoration, hep-th/0604161.
[10] R. Casero, A. Paredes, J. Sonnenschein, Fundamental matter, meson spectroscopy and non-critical string/gauge
duality, JHEP 0601 (2006) 127, hep-th/0510110.
[11] E. Antonyan, J.A. Harvey, S. Jensen, D. Kutasov, NJL and QCD from string theory, hep-th/0604017.
[12] K. Peeters, J. Sonnenschein, M. Zamaklar, Holographic melting and related properties of mesons in a quarkgluon
plasma, Phys. Rev. D 74 (2006) 106008, hep-th/0606195.
[13] J. Erlich, E. Katz, D.T. Son, M.A. Stephanov, QCD and a holographic model of hadrons, Phys. Rev. Lett. 95 (2005)
261602, hep-ph/0501128.
[14] L. Da Rold, A. Pomarol, Chiral symmetry breaking from five dimensional spaces, Nucl. Phys. B 721 (2005) 79,
hep-ph/0501218.
[15] J.P. Shock, F. Wu, Three flavour QCD from the holographic principle, JHEP 0608 (2006) 023, hep-ph/0603142;
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
133
C. Csaki, M. Reece, Toward a systematic holographic QCD: A braneless approach, hep-ph/0608266;

N. Evans, A. Tedder, Perfecting the ultra-violet of holographic descriptions of QCD, Phys. Lett. B 642 (2006) 546,
hep-ph/0609112;
J.P. Shock, F. Wu, Y.L. Wu, Z.F. Xie, AdS/QCD phenomenological models from a back-reacted geometry, hepph/0611227.
A. Karch, E. Katz, D.T. Son, M.A. Stephanov, Linear confinement and AdS/QCD, hep-ph/0602229.
U. Gursoy, E. Kiritsis, Exploring improved holographic theories for QCD: Part I, arXiv: 0707.1324 [hep-th];
U. Gursoy, E. Kiritsis, F. Nitti, Exploring improved holographic theories for QCD: Part II, arXiv: 0707.1349
[hep-th].
S. Sugimoto, K. Takahashi, QED and string theory, JHEP 0404 (2004) 051, hep-th/0403247.
F. Bigazzi, R. Casero, A.L. Cotrone, E. Kiritsis, A. Paredes, Non-critical holography and four-dimensional CFTs
with fundamentals, JHEP 0510 (2005) 012, hep-th/0505140.
A. Sen, Tachyon dynamics in open string theory, Int. J. Mod. Phys. A 20 (2005) 5513, hep-th/0410103.
Y. Kinar, E. Schreiber, J. Sonnenschein, Qanti-Q potential from strings in curved spacetime: Classical results,
Nucl. Phys. B 566 (2000) 103, hep-th/9811192;
J. Sonnenschein, What does the string/gauge correspondence teach us about Wilson loops?, hep-th/0003032.
M. Shifman, Highly excited hadrons in QCD and beyond, hep-ph/0507246.
S.R. Coleman, E. Witten, Chiral symmetry breakdown in large N chromodynamics, Phys. Rev. Lett. 45 (1980) 100.
S.L. Adler, Axial vector vertex in spinor electrodynamics, Phys. Rev. 177 (1969) 2426;
J.S. Bell, R. Jackiw, A PCAC puzzle: 0 in the sigma model, Nuovo Cimento A 60 (1969) 47.
I.R. Klebanov, J.M. Maldacena, Superconformal gauge theories and non-critical superstrings, Int. J. Mod. Phys.
A 19 (2004) 5003, hep-th/0409133.
A. Fotopoulos, V. Niarchos, N. Prezas, D-branes and SQCD in non-critical superstring theory, JHEP 0510 (2005)
081, hep-th/0504010;
S. Murthy, J. Troost, D-branes in non-critical superstrings and duality in N = 1 gauge theories with flavor,
JHEP 0610 (2006) 019, hep-th/0606203.
A. Sen, DiracBornInfeld action on the tachyon kink and vortex, Phys. Rev. D 68 (2003) 066008, hep-th/0303057.
M.R. Garousi, D-brane anti-D-brane effective action and brane interaction in open string channel, JHEP 0501 (2005)
029, hep-th/0411222.
E. Katz, A. Lewandowski, M.D. Schwartz, Tensor mesons in AdS/QCD, Phys. Rev. D 74 (2006) 086004, hepph/0510388.
D. Kutasov, M. Marino, G.W. Moore, Remarks on tachyon condensation in superstring field theory, hep-th/0010108.
J.A. Minahan, B. Zwiebach, Effective tachyon dynamics in superstring theory, JHEP 0103 (2001) 038, hepth/0009246.
A.A. Tseytlin, On non-Abelian generalisation of the BornInfeld action in string theory, Nucl. Phys. B 501 (1997)
41, hep-th/9701125.
M.R. Garousi, Tachyon couplings on non-BPS D-branes and DiracBornInfeld action, Nucl. Phys. B 584 (2000)
284, hep-th/0003122.
E. Witten, Anti-de Sitter space, thermal phase transition, and confinement in gauge theories, Adv. Theor. Math.
Phys. 2 (1998) 505, hep-th/9803131.
C. Vafa, E. Witten, Restrictions on symmetry breaking in vector-like gauge theories, Nucl. Phys. B 234 (1984) 173.
C. Kennedy, A. Wilkins, RamondRamond couplings on braneantibrane systems, Phys. Lett. B 464 (1999) 206,
hep-th/9905195.
P. Kraus, F. Larsen, Boundary string field theory of the D D system, Phys. Rev. D 63 (2001) 106004;
hep-th/0012198.
T. Takayanagi, S. Terashima, T. Uesugi, Braneantibrane action from boundary string field theory, JHEP 0103
(2001) 019, hep-th/0012210.
M.B. Green, J.A. Harvey, G.W. Moore, I-brane inflow and anomalous couplings on D-branes, Class. Quantum
Grav. 14 (1997) 47, hep-th/9605033.
D.Z. Freedman, S.D. Mathur, A. Matusis, L. Rastelli, Correlation functions in the CFT(d)/AdS(d + 1) correspondence, Nucl. Phys. B 546 (1999) 96, hep-th/9804058.
R. Jackiw, Topological investigations of quantized gauge theories, in: S.B. Treiman, R. Jackiw, B. Zumino, E.
Witten (Eds.), Current Algebra and Anomalies, World Scientific, 1985.
E. Schreiber, Excited mesons and quantization of string endpoints, hep-th/0403226.
M. Gell-Mann, R.J. Oakes, B. Renner, Behavior of current divergences under SU(3) SU(3), Phys. Rev. 175 (1968)
2195.
134
[44] A. Armoni, WittenVeneziano from GreenSchwarz, JHEP 0406 (2004) 019, hep-th/0404248.
[45] O. Bergman, G. Lifschytz, Holographic U(1)A and string creation, hep-th/0612289.
[46] J.M. Maldacena, C. Nunez, Towards the large N limit of pure N = 1 super-YangMills, Phys. Rev. Lett. 86 (2001)
588, hep-th/0008001.
[47] E. Witten, Theta dependence in the large N limit of four-dimensional gauge theories, Phys. Rev. Lett. 81 (1998)
2862, hep-th/9807109.
[48] E. Witten, Current algebra theorems for the U(1) Goldstone boson, Nucl. Phys. B 156 (1979) 269;
G. Veneziano, U(1) without instantons, Nucl. Phys. B 159 (1979) 213.
[49] L. Da Rold, A. Pomarol, The scalar and pseudoscalar sector in a five-dimensional approach to chiral symmetry
breaking, JHEP 0601 (2006) 157, hep-ph/0510268.
[50] S. Kuperstein, J. Sonnenschein, Non-critical supergravity (d > 1) and holography, JHEP 0407 (2004) 049, hepth/0403254;
S. Kuperstein, J. Sonnenschein, Non-critical, near extremal AdS(6) background as a holographic laboratory of fourdimensional YM theory, JHEP 0411 (2004) 026, hep-th/0411009.
[51] D. Mateos, R.C. Myers, R.M. Thomson, Thermodynamics of the brane, hep-th/0701132.
[52] R. Casero, C. Nunez, A. Paredes, Towards the string dual of N = 1 SQCD-like theories, Phys. Rev. D 73 (2006)
086005, hep-th/0602027;
A. Paredes, On unquenched N = 2 holographic flavor, JHEP 0612 (2006) 032, hep-th/0610270;
F. Benini, F. Canoura, S. Cremonesi, C. Nunez, A.V. Ramallo, Unquenched flavors in the KlebanovWitten model,
hep-th/0612118.
[53] K. Skenderis, Lecture notes on holographic renormalization, Class. Quantum Grav. 19 (2002) 5849, hep-th/0209067.
[54] A. Karch, A. OBannon, K. Skenderis, Holographic renormalization of probe D-branes in AdS/CFT, JHEP 0604
(2006) 015, hep-th/0512125.
Level-rank duality of untwisted and twisted D-branes

of the so(N)K WZW model
Stephen G. Naculich , Benjamin H. Ripman
Department of Physics Bowdoin College Brunswick, ME 04011, USA
Received 14 June 2007; accepted 11 July 2007
Abstract
We analyze the level-rank duality of untwisted and -twisted D-branes of the so(N )K WZW model. Untwisted D-branes of so(N )K are characterized by integrable tensor and spinor representations of so(N )K .
Level-rank duality maps untwisted so(N )K D-branes corresponding to (equivalence classes of ) tensor representations onto those of so(K)N . The -twisted D-branes of so(2n)2k are characterized by (a subset of )
integrable tensor and spinor representations of so(2n 1)2k+1 . Level-rank duality maps spinor -twisted
so(2n)2k D-branes onto those of so(2k)2n . For both untwisted and -twisted D-branes, we prove that the
spectrum of an open string ending on these D-branes is isomorphic to the spectrum of an open string ending
on the level-rank-dual D-branes.
1. Introduction
It has long been known that the modular transformation matrix and fusion algebra of the
WessZuminoWitten (WZW) model with affine Lie algebra su(N )K are closely related to those
of the WZW model with affine Lie algebra su(K)N (level-rank duality) [13]. Similar dualities
have been shown for WZW models with affine Lie algebras related by sp(n)k sp(k)n and
so(N )K so(K)N [2,3], and also for u (N )K,N(K+N ) u (K)N,K(K+N ) [4].
More recently, it has been shown [57] that the untwisted and twisted D-branes in the boundary su(N )K WZW model [825] respect level-rank duality; that is, there exists a one-to-one map
Research supported in part by the NSF under grant PHY-0456944.
E-mail address: naculich@bowdoin.edu (S.G. Naculich).

doi:10.1016/j.nuclphysb.2007.07.011
136
S.G. Naculich, B.H. Ripman / Nuclear Physics B 787 (2007) 135151
between the (equivalence classes of) D-branes of su(N )K and those of su(K)N . The open-string
spectra associated with level-rank-dual D-branes are isomorphic, and the charges of level-rankdual untwisted D-branes are equal (modulo sign), with a slightly more complicated relationship
holding between the charges of twisted D-branes. Level-rank duality also holds for the D-branes
of sp(n)k [6].
In this paper, we continue the story by establishing the level-rank duality of the untwisted
D-branes of so(N )K and of the twisted D-branes of so(2n)2k . In this case, level-rank duality is
partial and holds only for a subset of the D-branes of the theory. Moreover, in this case we find
no simple relation between the charges of level-rank-dual D-branes.
We begin by summarizing our results. Untwisted D-branes of so(N )K correspond to untwisted
Cardy states |C (boundary states of the bulk WZW model), which are labelled by integrable
highest-weight representations (both tensors and spinors) of the untwisted affine Lie algebra
so(N )K . Only untwisted tensor D-branes exhibit level-rank duality,1 and the duality is one-toone between equivalence classes [] of integrable tensor representations generated by the Z2 automorphisms and (when N is even) of the so(N )K algebra. (We denote by the simple
current symmetry of so(N )K that acts on the Dynkin indices of a representation by2 a0 a1 .
We denote by the chirality-flip symmetry of so(2n)K that acts on the Dynkin indices of a
representation by an an1 . For so(2n + 1)K , we define to be the identity.) The boundary
state corresponding to the equivalence class [] may be written as

1
[] =
|C + () C + () C + () C ,
t
()s()+3
2
where

1 if = (),
1 if = (),
t () =
s() =
0 if = ().
0 if = (),
(1.1)
(1.2)
Equivalence classes [] of integrable tensor representations of so(N )K are characterized by

Young tableaux with N/2 rows and K/2 columns. Level-rank duality acts by transposing
these tableaux, inducing a one-to-one correspondence [] []
between equivalence classes
of so(N )K and so(K)N , and therefore between the untwisted D-branes that correspond to the
boundary states (1.1). We show that the spectrum of representations carried by an open string
stretched between untwisted so(N )K D-branes corresponding to [] and [] is isomorphic to
that carried by an open string stretched between untwisted so(K)N D-branes corresponding to
[]
and [].
The so(2n)K WZW model contains, in addition to untwisted D-branes, a class of D-branes
twisted by the symmetry . These -twisted D-branes can be characterized [20] by (a subset of)
the integrable highest-weight representations (both tensors and spinors) of the untwisted affine
Lie algebra so(2n 1)K+1 . Only spinor -twisted D-branes of so(2n)2k exhibit level-rank duality, which involves a one-to-one map between the spinor -twisted D-branes of so(2n)2k
and the spinor -twisted D-branes of so(2k)2n . We show that the spectrum of representations
carried by an open string stretched between -twisted so(2n)2k D-branes corresponding to and
is isomorphic to that carried by an open string stretched between -twisted so(2k)2n D-branes
corresponding to and .
1 Except for s
o(2n + 1)2k+1 , where a level-rank map can also be defined between equivalence classes of untwisted
spinor D-branes.
2 Except for s
o(4)K , in which case acts by a0 min(a1 , a2 ).
137
This paper is organized as follows. Section 2 briefly reviews the Ishibashi and Cardy states of
the WZW model, and in Section 3, we characterize the integrable highest-weight representations
of so(N )K . Section 4 describes the level-rank duality of the (equivalence classes of) untwisted
D-branes corresponding to tensor representations of so(N )K and to spinor representations of
so(2n + 1)2k+1 . The -twisted Ishibashi and Cardy states of so(2n)K are reviewed in Section 5,
and in Section 6 we describe the level-rank duality of spinor -twisted D-branes of so(2n)2k .
2. Untwisted and twisted D-branes of WZW models
In this section, we briefly review some general aspects of untwisted and twisted D-branes of
the WZW model and their relation to the Cardy and Ishibashi states of the closed-string sector,
drawing on Refs. [9,10,18,20].
The WZW model, which describes strings propagating on a group manifold, is a rational
conformal field theory whose chiral algebra (for both left- and right-movers) is an untwisted
affine Lie algebra g K at level K. We only consider WZW theories with a diagonal closed-string
spectrum:
V V ,
Hclosed =
(2.1)
P+K
where V and V represent left- and right-moving states respectively, denotes the representation
conjugate to , and P+K is the set of integrable highest-weight representations of g K .
D-branes of the WZW model may be described algebraically in terms of the possible boundary
conditions that can consistently be imposed on a WZW model with boundary. We only consider
boundary conditions on the currents of the affine Lie algebra of the form
a

J (z) Ja (z) z=z = 0
(2.2)
where is an automorphism of the Lie algebra g. These boundary conditions leave unbroken the
g K symmetry, as well as the conformal symmetry, of the theory.
Twisted Ishibashi states
Open-closed string duality allows one to correlate the boundary conditions (2.2) of the boundary WZW model with coherent states |B Hclosed of the bulk WZW model satisfying
a

a
Jm + Jm
(2.3)
|B = 0, m Z,
where Jma are the modes of the affine Lie algebra generators. Solutions of Eq. (2.3) that belong
to a single sector V V() of the bulk WZW model are known as -twisted Ishibashi states
|I [26]. As we are considering the diagonal closed-string theory (2.1), -twisted Ishibashi
states only exist when = (), and so are labelled by E , where E is the subset of integrable representations of g K that satisfy () = . Equivalently, corresponds to an integrable
representation of g , the orbit Lie algebra associated with g K [27].
Twisted Cardy states
A coherent state |B that corresponds to an allowed boundary condition must also satisfy
additional (Cardy) conditions [28]. Solutions of Eq. (2.3) that satisfy the Cardy conditions are
138
denoted -twisted Cardy states |C , where the labels take values in B , the set of integrable
[10]. The -twisted Cardy states may be
representations of the -twisted affine Lie algebra g K
expressed as linear combinations of -twisted Ishibashi states
|C =

|I ,
S
0
E
(2.4)
where S is the modular transformation matrix of g K , 0 denotes the identity representation, and
the coefficients may be identified with the modular transformation matrices of the -twisted
[10].
affine Lie algebra g K
The -twisted D-branes of g K correspond to the -twisted Cardy states |C and are therefore also labelled by B . The spectrum of an open string stretched between -twisted
D-branes labelled by and is encoded in the open-string partition function

open
Z ( ) =
(2.5)
n ( ),
P+K
where ( ) is the affine character of the integrable highest-weight representation of g K . The

multiplicity n of the representation carried by the open string may be expressed as [20]
n =
S

E
S0
(2.6)
Untwisted Ishibashi and Cardy states

Untwisted Cardy states |C and untwisted Ishibashi states |I are solutions of Eq. (2.3)
with = 1, and both are labelled by integrable representations of g K . The matrix in
Eq. (2.4) relating the untwisted Cardy states to the untwisted Ishibashi states is given by the
modular transformation matrix S of g K [28]. Consequently, by virtue of Eq. (2.6) and the
Verlinde formula for the fusion coefficients [29]
n =
S S S
P+K
S0
= N ,
(2.7)
the multiplicities n of the representations carried by an open string stretched between two
untwisted D-branes and are given by the fusion coefficients N of the WZW model.
3. Integrable representations of so(N)K
In this section, we review some details about the integrable representations of so(N )K used
throughout this paper.3 Integrable representations
of an affine Lie algebra g K have non-negative

Dynkin indices (a0 , a1 , . . . , ar ) that satisfy ri=0 mi ai = K, where mi are the dual Coxeter labels
of the Dynkin diagram for g K , and r + 1 is the rank of g K .
3 Throughout this paper, N 3 is understood.
139
Integrable representations of so(2n + 1)K

The Dynkin diagram for so(2n + 1)K is
and the dual Coxeter labels are (m0 , m1 , m2 , . . . , mn1 , mn ) = (1, 1, 2, . . . , 2, 1), where the labelling of nodes is indicated on the diagram. Integrable representations of so(2n + 1)K thus have
Dynkin indices that satisfy4
a0 + a1 + 2(a2 + + an1 ) + an = K.
(3.1)
An even or odd value of an corresponds, respectively, to a tensor or spinor representation of

so(2n + 1). With each irreducible tensor representation of so(2n + 1) may be associated a Young
tableau whose row lengths i are given by
1
n1
j =i aj , for 1 i n 1,
2 an +
i =
(3.2)
1
for i = n.
2 an ,
The integrability condition (3.1) is equivalent to the constraint 1 + 2 K on the row lengths
of the tableau.
We may also formally use Eq. (3.2) to define row lengths for a spinor representation. These
row lengths are all half-integers, and correspond to a Young tableau containing a column of
half-boxes.
Integrable representations of so(2n)K
The Dynkin diagram for so(2n)K is
and the dual Coxeter labels are (m0 , m1 , m2 , . . . , mn2 , mn1 , mn ) = (1, 1, 2, . . . , 2, 1, 1), where
the labelling of nodes is indicated on the diagram. Integrable representations of so(2n)K thus
have Dynkin indices that satisfy5
a0 + a1 + 2(a2 + + an2 ) + an1 + an = K.
(3.3)
An even or odd value of an an1 corresponds, respectively, to a tensor or spinor representation

of so(2n).
4 Throughout this paper, by s
o(3)K we mean the affine Lie algebra su(2)2K . Its integrable representations have so(3)
Young tableaux that obey 1 K. Since 1 = 12 a1 , this means that Eq. (3.1) is replaced by a0 + a1 = 2K when n = 1.
5 For s
o(4)K , we take the integrability condition to be a0 + max(a1 , a2 ) = K, which is equivalent to 1 + |2 | K.
140
The Dynkin diagram of so(2n) (and also so(2n)K ) is invariant under the exchange of the
(n 1)th and nth nodes. This gives rise to a Z2 -automorphism of the so(2n) Lie algebra,
which exchanges representations with Dynkin indices (. . . , an1 , an ) and (. . . , an , an1 ). This
automorphism may be dubbed [20] chirality flip as it exchanges the two fundamental spinor
representations of opposite chirality.
For each representation of so(2n) we may define
(a + an1 ) + n2
j =i aj , for 1 i n 2,
2 n
1
i = 2 (an + an1 ),
(3.4)
for i = n 1,
1
for i = n,
2 (an an1 ),
in terms of which the integrability condition (3.3) becomes 1 + 2 K. The absolute values of
i represent the row lengths of a Young tableau A with up to n rows. (For spinor representations,
these row lengths are all half-integer, and correspond to a Young tableau containing a column of
half-boxes.) When an = an1 , the Young tableau has n 1 or fewer rows, and corresponds to a
unique irreducible so(2n) representation a, one which is invariant under . When an = an1 , the
Young tableau has precisely n rows and corresponds to two distinct representations, a and (a).
Hence we may consider the Young tableau A as labelling either an irreducible (a) or a reducible
(a (a)) representation of so(2n), depending respectively on whether the representation a is or
is not invariant under . We thus write

A = 2s(a)1 a (a) ,
(3.5)
where
s(a) =
1, if n = 0, that is, a = (a),

0, if n = 0, that is, a = (a).
(3.6)
(For all representations of so(2n + 1), we define (a) = a and s(a) = 0.)
Let Sab denote the (symmetric) modular transformation matrix of so(N )K , an explicit formula
for which may be found, for example, in Ref. [3]. We define
SAb = 2s(a)1 [Sab + S(a)b ],
SAB = 2s(b)1 [SAb + SA(b) ].
(3.7)
Since the modular transformation matrix obeys

S(a)b = Sa(b) ,
(3.8)
it follows that
SAb = SA(b) ,
SAB = 2s(b) SAb .
(3.9)
Simple current orbits of so(N )K

Both so(2n + 1)K and so(2n)K Dynkin diagrams have a Z2 -symmetry that exchanges the 0th
and 1st nodes. This symmetry induces a simple-current symmetry (denoted by ) of the so(N )K
WZW model that pairs integrable representations related by a0 a1 , with the other Dynkin
141
indices unchanged.6 Their respective Young tableaux are related by 1 K 1 . Under ,

tensor representations are mapped to tensors, and spinor representations to spinors.
We will refer to representations of so(N )K with 1 < 12 K, 1 = 12 K, and 1 > 12 K as being
of types I, II, and III, respectively. Type II representations are invariant under , and are tensors
(resp. spinors) when K is even (resp. odd). Each simple-current orbit of so(N )K contains either
a type I and type III representation, or a single type II representation. We define

1, if 1 = 12 K (type II), that is, a = (a),
t (a) =
(3.10)
0, if 1 = 12 K (type I or III), that is, a = (a).
Finally, the modular transformation matrix of so(N )K obeys

tensor
S (a)b = S(a)b , for b a
representation.
spinor
(3.11)
4. Level-rank duality of untwisted D-branes of so(N)K

Having reviewed the characterization of integrable representations of so(N )K in the previous
section, we now turn to the untwisted D-branes of the so(N )K WZW model, which are labelled
by those representations. In this section, we will demonstrate a level-rank duality between the
untwisted D-branes of so(N )K and those of so(K)N .
Since the multiplicities of the representations carried by an open string stretched between
two untwisted D-branes are given by the fusion coefficients of the WZW model (2.7), level-rank
duality of the untwisted D-branes of the so(N )K model is closely related to level-rank duality
of the fusion coefficients of this model, which was described in Ref. [3]. We recall two salient
aspects of this duality:
The level-rank map is partial: it only relates the tensor representations7 of so(N )K to those
of so(K)N .
The level-rank map is not one-to-one between integrable tensor representations a, but rather
between equivalence classes of representations,8 denoted by [a]. These equivalence classes
are characterized by tensor Young tableaux with N/2 rows and K/2 columns (termed
reduced and cominimally-reduced in Ref. [3]). Level-rank duality acts by transposing
these tableaux, and maps the set of tensor Young tableaux with N/2 rows and K/2
columns one-to-one onto the set of tensor Young tableaux with K/2 rows and N/2
columns.
The equivalence classes of integrable tensor representations fall into several categories, which
we now describe, using the notation of the previous section.
(1) s(a) = 0 and t (a) = 1: the equivalence class labelled by a tensor Young tableau with 1 =
1
1
2 K columns (only possible when K is even) and with fewer than 2 N rows corresponds to a
6 Except for s
o(4)K , in which case acts by a0 min(a1 , a2 ). Thus, if a has Dynkin indices (a0 , a1 , a2 ) then (a)
has Dynkin indices (min(a1 , a2 ), K a2 , K a1 ).
7 For s
o(2n + 1)2k+1 , a level-rank map between the spinor representations also exists [3], and thus a level-rank map
can be defined for all the untwisted D-branes of so(2n + 1)2k+1 .
8 This is also the case for level-rank duality of s
u(N )K .
142
single (type II) irreducible representation a, whose Dynkin indices satisfy a0 = a1 and (for
N = 2n) an = an1 . This representation is invariant under both and .
(2) s(a) = 0 and t (a) = 0: the equivalence class labelled by a tensor Young tableau with
1 < 12 K columns and with fewer than 12 N rows corresponds to a pair of irreducible representations a and (a) (of type I and type III) whose Dynkin indices are related by a0 a1 .
When N = 2n, the Dynkin indices of these representations satisfy an = an1 , i.e., these
representations are invariant under .
(3) s(a) = 1 and t (a) = 1: the equivalence class labelled by a tensor Young tableau with 1 =
1
1
2 K columns (only possible when K is even) and with exactly 2 N rows (only possible when
N is even) corresponds to a pair of (type II) irreducible representations a and (a), whose
Dynkin indices are related by an an1 where N = 2n, and obey9 a0 = a1 , i.e., these
representations are invariant under .
(4) s(a) = 1 and t (a) = 0: the equivalence class labelled by a tensor Young tableau with 1 <
1
1
2 K columns and with exactly 2 N rows (only possible when N is even) corresponds to four
irreducible representations: a, (a), (a), and ( (a)) (two of type I and two of type III).
Let [a]
denote the transpose of the Young tableau characterizing the equivalence class [a]. Then
t (a) = s (a)
and s(a) = t(a),
(4.1)
where s and t are the quantities (3.6) and (3.10) defined for so(K)N . Under level-rank duality,
equivalence classes [a] in categories (1), (2), (3), and (4) map into equivalence classes [a]
in
categories (4), (2), (3), and (1), respectively.
We now elucidate the implications of level-rank duality for the untwisted tensor D-branes of
the so(N )K WZW model. Consider the linear combination of untwisted Cardy states

1
[] =
|C + () C + () C + () C ,
2 t ()s()+3
(4.2)
which corresponds to an equivalence class [] of integrable tensor representations. Using

Eqs. (2.7) and (3.11), we find that the multiplicity n[][] [] of the equivalence class of representations [] carried by an open string stretched between untwisted D-branes corresponding to
the states |[] and [] is given by
n[][] [] =
1
2 t ()+t ()+t ()s()s()s()+3
+ S
(S + S() )(S + S() )(S
() )
=tensor
representations
S0
(4.3)
where only integrable tensor representations remain in the sum as a consequence of Eq. (3.11).
Using Eqs. (3.7) and (3.9), we express n[][] [] in terms of Young tableaux A, B, , M, related
to , , , and by Eq. (3.5),
9 For s
o(4)K , they obey a0 = min(a1 , a2 ).
1
n[][] [] =
t
()+t
()+t
()+s()+s()+s()3
2

1 SBM SM SAM
.
S0M
2s()
143
(4.4)
M=tensor
tableaux I, II, III
Finally, since SA (M) = SAM , the sum may be restricted to tableaux of types I and II
(cominimally-reduced tableaux)
1
n[][] [] = t ()+t ()+t ()+s()+s()+s()3
2

M=tensor
tableaux I, II
SBM SM SAM
.
S0M
2s()+t ()1
(4.5)
The multiplicities n[][] [] are closely related to B A defined in Eq. (3.16) of Ref. [3].
[]
Level-rank duality maps the state |[] of so(N )K to the state |[]
of so(K)N . Let n [][
]
denote the quantity (4.3) defined for so(K)N . The form of Eq. (4.5) makes manifest the equality
of the multiplicities
[]
n[][] [] = n [][
]
(4.6)
as a consequence of three facts: (1) the set of cominimally-reduced tableaux M of so(N )K are
in one-to-one correspondence with those of so(K)N , (2) Eq. (4.1) holds for all tensor representations, and (3) the quantities SAB , defined by Eq. (3.7), are level-rank dual (SAB = SA B ) as was
proved in the appendix of Ref. [3]. Hence, the spectrum of representations carried by open strings
stretched between untwisted tensor D-branes of so(N )K is level-rank dual.
We end this section by describing the level-rank duality of untwisted spinor D-branes of
so(2n + 1)2k+1 . The equivalence classes [] of spinor representations of so(2n + 1)2k+1 are
characterized by type I and type II spinor tableaux, where a type I tableau represents a pair of
spinor representations and (), and a type II tableau represents a single irreducible spinor representation that obeys () = . The level-rank map [] []
between equivalence classes
of spinor representations of so(2n + 1)2k+1 and so(2k + 1)2n+1 was presented in Ref. [3]:
reduce each of the row lengths by 12 , so that they all become integers,
transpose the resulting tableau,
take the complement with respect to a k n rectangle and rotate 180 degrees,
add 12 to each of the row lengths.
This takes type I spinor tableaux of so(2n + 1)2k+1 to type II spinor tableaux of so(2k + 1)2n+1
and vice versa: t () = 1 t().
This procedure thus defines a map between an untwisted spinor
D-brane of so(2n + 1)2k+1 corresponding to the boundary state

[] = 1
|C + () C
2 t ()+1
(4.7)
and an untwisted spinor D-brane |[]
of so(2k + 1)2n+1 . The multiplicity of the (equivalence class of) representations [] carried by an open string stretched between untwisted spinor
144
D-branes [] and [] of so(2n + 1)2k+1 obeys

n[][] [] =
1
2 t ()+t ()5
S S S
=tensor
tableau I
S0
[]
= n [][
] ,
(4.8)
using Eq. (3.25) of Ref. [3]. Hence, the spectrum of representations carried by open strings
stretched between untwisted spinor D-branes of so(2n + 1)2k+1 is also level-rank dual.
5. The -twisted D-branes of the so(2n)K model
In the previous section, we proved the level-rank duality of untwisted D-branes of so(N )K . In
this section, we will describe a class of twisted D-branes of so(2n)K , and in the next section, we
will prove the level-rank duality of a subset of these twisted D-branes.
Recall from in Section 3 that the finite Lie algebra so(2n) possesses (when n 2) a Z2 automorphism (chirality flip), under which the Dynkin indices an1 and an of an irreducible
representation are exchanged. This automorphism lifts to an automorphism of the affine Lie
algebra so(2n)K , and gives rise to a set of -twisted Ishibashi states and -twisted Cardy states of
the bulk so(2n)K WZW model, and a corresponding class of -twisted D-branes of the boundary
model. In this section we characterize these twisted states, relying heavily on Ref. [20].
-twisted Ishibashi states of so(2n)K
The -twisted Ishibashi states |I of the so(2n)K WZW model are labelled by integrable
(1)
representations E of so(2n)K = (Dn )K that obey () = (i.e., integrable representations characterized by so(2n) tensor Young tableaux with no more than n 1 rows). These
representations have Dynkin indices
(0 , 1 , . . . , n2 , n1 , n1 ),
that
(5.1)
satisfy10
0 + 1 + 2(2 + + n1 ) = K.
(5.2)
Equivalently, the -twisted Ishibashi states of so(2n)K may be characterized by the integrable
representations of the associated orbit Lie algebra g = (A(2)
2n3 )K [20] whose Dynkin diagram
is
and whose dual Coxeter numbers are (m0 , m1 , m2 , . . . , mn1 ) = (1, 1, 2, . . . , 2), where the la(1)
belling of nodes is indicated on the diagram. The (Dn )K representation with Dynkin indices
(2)
(5.1) corresponds to the (A2n3 )K representation with Dynkin indices (0 , 1 , . . . , n2 , n1 ),
whose integrability condition is precisely (5.2).
10 For s
o(4)K , the Dynkin indices (0 , 1 , 1 ) satisfy 0 + 1 = K.
145
It was shown in Ref. [20] that each -twisted Ishibashi state of so(2n)K may be mapped
to a spinor representation of the untwisted affine Lie algebra so(2n 1)K+1 with Dynkin
indices11
i = i
(0 i n 2) and
n1 = 2n1 + 1.
(5.3)
The constraint (5.2) on is precisely equivalent to the integrability constraint (3.1) on the representation of so(2n 1)K+1 . Hence, -twisted Ishibashi states of so(2n)K are in one-to-one
correspondence with the set of integrable spinor representations of so(2n 1)K+1 of type I, type
II (when K is even), and type III.
-twisted Cardy states of so(2n)K
The -twisted Cardy states |C (and therefore the -twisted D-branes) of the so(2n)K WZW
model12 are labelled by the integrable representations B of the -twisted affine Lie algebra
= (D (2) ) [20], whose Dynkin diagram is
g K
n K
and whose dual Coxeter numbers are (m0 , m1 , . . . , mn2 , mn1 ) = (1, 2, . . . , 2, 1), where the
labelling of nodes is indicated on the diagram. The Dynkin indices (0 , 1 , . . . , n2 , n1 ) of
thus satisfy13
0 + 2(1 + + n2 ) + n1 = K.
(5.4)
It was shown in Ref. [20] that each -twisted Cardy state of so(2n)K may be mapped to a
representation of the untwisted affine Lie algebra so(2n 1)K+1 with Dynkin indices14
0 = 0 + 1 + 1
and i = i
(1 i n 1).
implies that
(5.5)
is an integrable representation of so(2n 1)K+1 , and the

The constraint (5.4) on
constraint 0 0 further implies that is a type I representation of so(2n 1)K+1 (i.e., corresponds to a Young tableau whose first row length obeys 1 12 K). Therefore, -twisted D-branes
of so(2n)K are in one-to-one correspondence with the set of integrable type I representations of
so(2n 1)K+1 , of both tensor and spinor types.
Although the -twisted Ishibashi states and the -twisted Cardy states of so(2n)K are characterized differently in terms of integrable representations
of so(2n
n2+K/2
1)K+1 , they are equal in
number. (For K even, the number of each is n1+K/2
+
, while for K odd, the number
n1
n1
n1+(K1)/2
of each is 2
.) Thus, the -twisted Cardy states may be written as linear combin1
nations (2.4) of -twisted Ishibashi states , with the transformation coefficients given by
the modular transformation matrix of (Dn(2) )K . In Ref. [20], it was shown that these coefficients
are proportional to the (real) matrix elements S of the modular transformation matrix of the
11 Except for s
o(4)K , in which case 0 = 20 + 1 and 1 = 21 + 1. Also, recall footnote 4.
12 n 2 is understood.
13 For s
o(4)K , the condition is 0 + 1 = K.
14 Except for s
o(4)K , in which case 0 = 20 + 1 + 2 and 1 = 1 . Also, recall footnote 4.
146
untwisted affine Lie algebra so(2n 1)K+1 :
= 2S = 2S
(5.6)
where and are the so(2n 1)K+1 representations related to and by Eqs. (5.5) and (5.3),
respectively.
Twisted open string partition function of so(2n)K
Combining Eqs. (2.6) and (5.6), we may write the multiplicities of the representations carried
by an open string stretched between -twisted D-branes and of so(2n)K as
2S S S
n =
S0
=spinors I,II,III
(5.7)
where S and S are modular transformation matrix elements of so(2n)K and so(2n 1)K+1
respectively, and the sum is over all -twisted Ishibashi states of so(2n)K , or equivalently, over
all integrable spinor representations of so(2n 1)K+1 . (Type II spinors of so(2n 1)K+1 are
present only when K is even.)
Although the -twisted D-branes correspond to both tensor and spinor representations of
so(2n 1)K+1 , for the remainder of this section we will restrict and to correspond to spinor
representations and of so(2n 1)K+1 , which allows us to simplify Eq. (5.7) considerably.
Recall from Eq. (3.11) that the modular transformation matrix elements of so(2n 1)K+1 obey
S ( ) = S ,
for = spinor.
(5.8)
As a consequence, type I and III representations , which are related by , may be combined in
Eq. (5.7)
2S S S
2S S () S
n =
+
, for , both spinors (5.9)
S
S
0
0
()

=spinors I
and type II representations, which obey ( ) = , drop out of the sum since S = 0. (We
have also used the fact that the so(2n 1)K+1 representation ( ), related to by 1
K + 1 1 , corresponds via the map (5.3) to the -twisted Ishibashi state () of so(2n)K ,
related to by 1 K 1 .) Finally, recalling that the modular transformation matrix elements
of so(2n)K obey

tensor
representation
S () = S() , for a
(5.10)
spinor
and that -twisted Ishibashi states obey () = , we finally obtain
4S S S
n =
=spinors
S0
for , both spinors and = tensor,
(5.11)
and
n = 0,
for , both spinors and = spinor.
(5.12)
This result will allow us to demonstrate in the next section the level-rank duality of the spectrum
of an open string stretched between spinor -twisted D-branes of so(2n)2k .
147
6. Level-rank duality of -twisted D-branes of so(2n)2k

As we saw in the previous section, the so(2n)K WZW model possesses twisted D-branes
corresponding to the chirality-flip symmetry of the so(2n) Dynkin diagram, and these
-twisted D-branes are characterized by integrable type I tensor and spinor representations of
so(2n 1)K+1 . We will refer to these as tensor and spinor -twisted D-branes, respectively.
In this section, we will exhibit a level-rank duality15 between the -twisted D-branes of
so(2n)2k and those of so(2k)2n . This duality is partial, and only holds between spinor -twisted
D-branes (just as the level-rank duality of untwisted D-branes only holds between tensor
D-branes). The restriction to spinor -twisted D-branes
can be anticipated by observing that the
and the number of spinor -twisted
number of tensor -twisted D-branes of so(2n)2k is n+k1
n1
n+k2
D-branes is n1 , and only the latter is invariant under n k.
First we define an explicit one-to-one map between the spinor -twisted D-branes of
so(2n)2k and so(2k)2n . The map is defined by specifying its action16 on the corresponding
so(2n 1)2k+1 and so(2k 1)2n+1 representations and :
reduce each of the row lengths of by 12 , so that they all become integers,
transpose the resulting tableau,
add 12 to each of the row lengths.
The same procedure defines a one-to-one map between type I spinor representations
of so(2n 1)2k+1 and so(2k 1)2n+1 corresponding to -twisted Ishibashi states. By virtue of
Eq. (5.3), this map lifts to a map between (a subset of) the -twisted Ishibashi states. As
suggested by the notation, this map is simply transposition of the type I so(2n)2k Young tableau
corresponding to .
Next, we turn to the level-rank duality of the spectrum of an open string stretched between
-twisted D-branes. In the previous section, it was shown that the multiplicity of the representation carried by an open string stretched between spinor -twisted D-branes and of so(2n)2k
is given by
n =
4S S S
=spinors I
S0
for = tensor,
(6.1)
with n vanishing for = spinor. As in Section 4, however, we consider the multiplicity corresponding to the equivalence class of tensor representations []:

1
n + n() + n () + n( ()) .
n[] =
2 t ()s()+3
Using Eqs. (6.1), (3.11), and (3.7), we find
1
n[] =
t
()s()+1
2

=spinors I
(6.2)
4S (S + S() )S
S0
15 Clearly the -twisted D-branes of s

o(2n)2k+1 have no level-rank duals, since so(2k + 1)2n has no -twisted D-branes.
16 Note that the hat map defined here differs from that defined in Section 4 between spinor representations of
so(2n + 1)2k+1 and so(2k + 1)2n+1 . The map defined here also characterizes the map between c -twisted D-branes
of su(2n + 1)2k+1 and su(2k + 1)2n+1 [7].
148
=
2 t ()+s()1
4S SM S

=spinors I
S0M
(6.3)
where = 2s()1 [ ()] and M = since () = . The form of Eq. (6.3) makes manifest
the equality of the multiplicities
1
n[] =
t
()+s()1
2
1
=
s
)+
t( )1
2
4S SM S

=spinors
4S S M S

=spinors
S0M
S0M
= n [
] ,
(6.4)
(6.5)
(6.6)
where we have used Eq. (4.1) and the facts that:

(1) type I spinors of so(2n 1)2k+1 map one-to-one to type I spinors of so(2k 1)2n+1 ,
(2) SM = S M [3], where S and S are the modular transformation matrices of so(2n)2k and
so(2k)2n respectively, and
(3) S = S for and both type I spinor representations (Eq. (6.10) of Ref. [7]), where
S and S are the modular transformation matrices of so(2n 1)2k+1 and so(2k 1)2n+1
respectively. Since by Eq. (5.12) only tensor representations appear in the -twisted openstring partition function (2.5), we have established that the spectrum of representations
carried by open strings stretched between -twisted D-branes of so(2n)2k is level-rank dual.
7. Conclusions
We have analyzed the level-rank duality of the untwisted D-branes of so(N )K and of the
-twisted D-branes of so(2n)2k . In each case, only a subset of the D-branes are mapped onto
those of the level-rank-dual theory.
Untwisted D-branes of so(N )K are characterized by integrable tensor and spinor representations of so(N )K . Only the untwisted tensor D-branes participate in level-rank duality.17
The tensor representations of so(N )K fall into equivalence classes [] generated by the
Z2 -isomorphisms and (the latter non-trivial only for N even), and characterized by Young
tableaux with N/2 rows and K/2 columns. Level-rank duality acts by transposing these
tableaux, and thus maps the equivalence classes [] of untwisted tensor D-branes of so(N )K onto
[]
of so(K)N . We showed that the multiplicity n[][] [] of the (equivalence class of) representations [] carried by an open string stretched between untwisted so(N )K D-branes corresponding
[]
to [] and [] is equal to n [][

] , the multiplicity of the (equivalence class of) representations
carried by an open string stretched between untwisted so(K)N D-branes corresponding to
[]
A similar result was shown for untwisted spinor D-branes of so(2n + 1)2k+1 .
[]
and [].
The -twisted D-branes of so(2n)2k , associated with the chirality-flip symmetry of the
so(2n) Dynkin diagram, are characterized by type I integrable tensor and spinor representations of so(2n 1)2k+1 . Only the spinor -twisted D-branes participate in level-rank duality.
17 Except for s
o(2n + 1)2k+1 , where the untwisted spinor D-branes also respect level-rank duality.
149
We defined a one-to-one map between the spinor -twisted D-branes of so(2n)2k and
the spinor -twisted D-branes of so(2k)2n . We then showed that the multiplicity n[] of the
(equivalence class of) representations [] carried by an open string stretched between -twisted
so(2n)2k D-branes corresponding to and is equal to n [

] , the multiplicity of the (equivalence class of) representations [ ] carried by an open string stretched between -twisted so(2k)2n
D-branes corresponding to and .

Hence, for both untwisted and -twisted D-branes, we have established an isomorphism between the spectrum of an open string ending on these D-branes and the spectrum of an open
string ending on the level-rank-dual D-branes.
In both the su(N )K and sp(n)k WZW theories, the charges of level-rank-dual untwisted
D-branes are equal (modulo sign) [5,6], with a slightly more complicated relationship holding
between the charges of twisted D-branes [7]. In the case of so(N )K , however, the charges of the
D-branes do not exhibit any simple relationship under level-rank duality.
Acknowledgements
S.N. wishes to thank Howard Schnitzer for his collaboration on a long series of papers on
which the results of this paper depend.
References
[1] S.G. Naculich, H.J. Schnitzer, Duality between SU(N )k and SU(k)N WZW models, Nucl. Phys. B 347 (1990)
687742;
S.G. Naculich, H.J. Schnitzer, Duality relations between SU(N )k and SU(k)N WZW models and their braid matrices, Phys. Lett. B 244 (1990) 235240;
M.A. Walton, Conformal branching rules and modular invariants, Nucl. Phys. B 322 (1989) 775;
D. Altschuler, M. Bauer, C. Itzykson, The branching rules of conformal embeddings, Commun. Math. Phys. 132
(1990) 349364;
J. Fuchs, P. van Driel, Some symmetries of quantum dimensions, J. Math. Phys. 31 (1990) 17701775;
A. Kuniba, T. Nakanishi, Level rank duality in fusion RSOS models, in: Proceedings of the International Colloquium
on Modern Quantum Field Theory, Bombay, January 1990, World Scientific, Singapore, 1991;
H. Saleur, D. Altschuler, Level rank duality in quantum groups, Nucl. Phys. B 354 (1991) 579613;
T. Nakanishi, A. Tsuchiya, Level rank duality of WZW models in conformal field theory, Commun. Math. Phys. 144
(1992) 351372.
[2] S.G. Naculich, H.A. Riggs, H.J. Schnitzer, Group level duality in WZW models and ChernSimons theory, Phys.
Lett. B 246 (1990) 417422.
[3] E.J. Mlawer, S.G. Naculich, H.A. Riggs, H.J. Schnitzer, Group level duality of WZW fusion coefficients and Chern
Simons link observables, Nucl. Phys. B 352 (1991) 863896.
[4] S.G. Naculich, H.J. Schnitzer, Level-rank duality of the U(N ) WZW model, ChernSimons theory, and 2d qYM
theory, JHEP 0706 (2007) 023, hep-th/0703089.
[5] S.G. Naculich, H.J. Schnitzer, Level-rank duality of D-branes on the SU(N ) group manifold, Nucl. Phys. B 740
(2006) 181194, hep-th/0511083.
[6] S.G. Naculich, H.J. Schnitzer, Level-rank duality of untwisted and twisted D-branes, Nucl. Phys. B 742 (2006)
295311, hep-th/0601175.
[7] S.G. Naculich, H.J. Schnitzer, Twisted D-branes of the SU(N )K WZW model and level-rank duality, Nucl. Phys.
B 755 (2006) 164185, hep-th/0606147.
[8] C. Klimcik, P. Severa, Open strings and D-branes in WZNW models, Nucl. Phys. B 488 (1997) 653676, hep-th/
9609112;
M. Kato, T. Okada, D-branes on group manifolds, Nucl. Phys. B 499 (1997) 583595, hep-th/9612148;
A.Y. Alekseev, V. Schomerus, D-branes in the WZW model, Phys. Rev. D 60 (1999) 061901, hep-th/9812193;
K. Gawedzki, Conformal field theory: A case study, hep-th/9904145.
150
[9] R.E. Behrend, P.A. Pearce, V.B. Petkova, J.-B. Zuber, On the classification of bulk and boundary conformal field
theories, Phys. Lett. B 444 (1998) 163166, hep-th/9809097;
J. Fuchs, C. Schweigert, Symmetry breaking boundaries. I: General theory, Nucl. Phys. B 558 (1999) 419483,
hep-th/9902132;
R.E. Behrend, P.A. Pearce, V.B. Petkova, J.-B. Zuber, Boundary conditions in rational conformal field theories,
Nucl. Phys. B 570 (2000) 525589, hep-th/9908036.
[10] L. Birke, J. Fuchs, C. Schweigert, Symmetry breaking boundary conditions and WZW orbifolds, Adv. Theor. Math.
Phys. 3 (1999) 671726, hep-th/9905038.
[11] A.Y. Alekseev, A. Recknagel, V. Schomerus, Non-commutative world-volume geometries: Branes on SU(2) and
fuzzy spheres, JHEP 9909 (1999) 023, hep-th/9908040;
A.Y. Alekseev, A. Recknagel, V. Schomerus, Brane dynamics in background fluxes and non-commutative geometry,
JHEP 0005 (2000) 010, hep-th/0003187;
A.Y. Alekseev, A. Recknagel, V. Schomerus, Open strings and non-commutative geometry of branes on group
manifolds, Mod. Phys. Lett. A 16 (2001) 325336, hep-th/0104054;
A. Alekseev, V. Schomerus, RR charges of D2-branes in the WZW model, hep-th/0007096.
[12] G. Felder, J. Frohlich, J. Fuchs, C. Schweigert, The geometry of WZW branes, J. Geom. Phys. 34 (2000) 162190,
hep-th/9909030.
[13] S. Stanciu, D-branes in group manifolds, JHEP 0001 (2000) 025, hep-th/9909163;
S. Stanciu, A note on D-branes in group manifolds: Flux quantization and D0-charge, JHEP 0010 (2000) 015,
hep-th/0006145;
S. Stanciu, An illustrated guide to D-branes in SU(3), hep-th/0111221;
J.M. Figueroa-OFarrill, S. Stanciu, D-brane charge, flux quantization and relative (co)homology, JHEP 0101 (2001)
006, hep-th/0008038.
[14] C. Bachas, M.R. Douglas, C. Schweigert, Flux stabilization of D-branes, JHEP 0005 (2000) 048, hep-th/0003037;
J. Pawelczyk, JHEP 0008 (2000) 006, hep-th/0003057;
W. Taylor, D2-branes in B fields, JHEP 0007 (2000) 039, hep-th/0004141.
[15] S. Fredenhagen, V. Schomerus, Branes on group manifolds, gluon condensates, and twisted K-theory, JHEP 0104
(2001) 007, hep-th/0012164.
[16] J.M. Maldacena, G.W. Moore, N. Seiberg, Geometrical interpretation of D-branes in gauged WZW models,
JHEP 0107 (2001) 046, hep-th/0105038;
J.M. Maldacena, G.W. Moore, N. Seiberg, D-brane instantons and K-theory charges, JHEP 0111 (2001) 062, hepth/0108100.
[17] K. Gawedzki, Boundary WZW, G/H, G/G and CS theories, Annal. Henri Poincare 3 (2002) 847881, hepth/0108044;
K. Gawedzki, N. Reis, WZW branes and gerbes, Rev. Math. Phys. 14 (2002) 12811334, hep-th/0205233.
[18] H. Ishikawa, Boundary states in coset conformal field theories, Nucl. Phys. B 629 (2002) 209232, hep-th/0111230.
[19] V.B. Petkova, J.B. Zuber, Boundary conditions in charge conjugate sl(N ) WZW theories, hep-th/0201239.
[20] M.R. Gaberdiel, T. Gannon, Boundary states for WZW models, Nucl. Phys. B 639 (2002) 471501, hep-th/0202067.
[21] M.R. Gaberdiel, T. Gannon, The charges of a twisted brane, JHEP 0401 (2004) 018, hep-th/0311242;
M.R. Gaberdiel, T. Gannon, D. Roggenkamp, The D-branes of SU(n), JHEP 0407 (2004) 015, hep-th/0403271;
M.R. Gaberdiel, T. Gannon, D. Roggenkamp, The coset D-branes of SU(n), JHEP 0410 (2004) 047, hepth/0404112.
[22] A.Y. Alekseev, S. Fredenhagen, T. Quella, V. Schomerus, Non-commutative gauge theory of twisted D-branes,
Nucl. Phys. B 646 (2002) 127157, hep-th/0205123;
T. Quella, Branching rules of semi-simple Lie algebras using affine extensions, J. Phys. A 35 (2002) 37433754,
math-ph/0111020;
T. Quella, V. Schomerus, Symmetry breaking boundary states and defect lines, JHEP 0206 (2002) 028, hepth/0203161;
T. Quella, On the hierarchy of symmetry breaking D-branes in group manifolds, JHEP 0212 (2002) 009, hepth/0209157.
[23] P. Bouwknegt, P. Dawson, D. Ridout, D-branes on group manifolds and fusion rings, JHEP 0212 (2002) 065, hepth/0210302;
P. Bouwknegt, D. Ridout, A note on the equality of algebraic and geometric D-brane charges in WZW models,
JHEP 0405 (2004) 029, hep-th/0312259.
[24] H. Ishikawa, T. Tani, Novel construction of boundary states in coset conformal field theories, Nucl. Phys. B 649
(2003) 205242, hep-th/0207177;
[25]
[26]
[27]
[28]
[29]
151
H. Ishikawa, A. Yamaguchi, Twisted boundary states in c = 1 coset conformal field theories, JHEP 0304 (2003)
026, hep-th/0301040;
H. Ishikawa, T. Tani, Twisted boundary states in KazamaSuzuki models, Nucl. Phys. B 678 (2004) 363397,
hep-th/0306227;
H. Ishikawa, T. Tani, Twisted boundary states and representation of generalized fusion algebra, hep-th/0510242;
S. Schafer-Nameki, D-branes in N = 2 coset models and twisted equivariant K-theory, hep-th/0308058;
V. Braun, S. Schafer-Nameki, Supersymmetric WZW models and twisted K-theory of SO(3), hep-th/0403287.
M.R. Gaberdiel, T. Gannon, D-brane charges on non-simply connected groups, JHEP 0404 (2004) 030, hepth/0403011;
S. Fredenhagen, D-brane charges on SO(3), JHEP 0411 (2004) 082, hep-th/0404017;
M. Vasudevan, Charges of exceptionally twisted branes, JHEP 0507 (2005) 035, hep-th/0504006;
S. Fredenhagen, M.R. Gaberdiel, T. Mettler, Charges of twisted branes: The exceptional cases, JHEP 0505 (2005)
058, hep-th/0504007;
S. Fredenhagen, T. Quella, Generalised permutation branes, JHEP 0511 (2005) 004, hep-th/0509153.
N. Ishibashi, The boundary and crosscap states in conformal field theories, Mod. Phys. Lett. A 4 (1989) 251.
J. Fuchs, B. Schellekens, C. Schweigert, From Dynkin diagram symmetries to fixed point structures, Commun.
Math. Phys. 180 (1996) 3998, hep-th/9506135.
J.L. Cardy, Boundary conditions, fusion rules and the Verlinde formula, Nucl. Phys. B 324 (1989) 581.
E.P. Verlinde, Fusion rules and modular transformations in 2d conformal field theory, Nucl. Phys. B 300 (1988)
360.
Cosmology and astrophysics of minimal dark matter

Marco Cirelli a,b, , Alessandro Strumia c , Matteo Tamburini d
a Service de Physique Thorique, CEA-Saclay, France
b INFN, Italy
c Dipartimento di Fisica dellUniversit di Pisa and INFN, Italy
d Dipartimento di Fisica dellUniversit di Pisa, Italy
Received 2 July 2007; accepted 20 July 2007

Available online 1 August 2007
Abstract
We consider DM that only couples to SM gauge bosons and fills one gauge multiplet, e.g., a fermion
5-plet (which is automatically stable), or a wino-like 3-plet. We revisit the computation of the cosmological
relic abundance including non-perturbative corrections. The predicted mass of, e.g., the 5-plet increases
from 4.4 to 10 TeV, and indirect detection rates are enhanced by 2 orders of magnitude. Next, we show
that due to the quasi-degeneracy among neutral and charged components of the DM multiplet, a significant
fraction of DM with energy E 1017 eV (possibly present among ultra-high energy cosmic rays) can cross
the Earth exiting in the charged state and may in principle be detected in neutrino telescopes.
1. Introduction
Observations and experiments tell that:
(i) Dark Matter (DM) exists [1], with abundance DM h2 = 0.110 0.005 [2];
(ii) DM is cosmologically stable; it has no electric charge [3], no strong interactions [4] and the
DM coupling to the Z boson is smaller than 103 g2 [5].
(i) suggests that DM might be a weak-scale particle: indeed DM 1 is the typical freezeout abundance of a typical particle with weak-scale mass M and coupling g: M/g
* Corresponding author at: Service de Physique Thorique, CEA-Saclay, France.
E-mail address: marco.cirelli@cea.fr (M. Cirelli).

doi:10.1016/j.nuclphysb.2007.07.023
M. Cirelli et al. / Nuclear Physics B 787 (2007) 152175
153
T0 MPl TeV. (ii) can be satisfied in appropriate models, although these properties are not
typical of weak scale particles.
We here want to study Minimal DM (MDM) models: we add to the Standard Model (SM) one
new multiplet, that only has gauge interactions and an SU(2)L -invariant mass M. In this context,
(ii) singles out scalars or fermions with zero hypercharge that fill an SU(2)L representation with
odd dimension: n = 3, 5, . . . [6]. The neutral DM0 component is then lighter than the charged
DM components, by only 166 MeV. DM0 interacts with ordinary matter via one-loop exchange
of W : future direct DM searches will hopefully reach the sensitivity needed to detect it.
The fermion 5-plet is particularly interesting: all interactions other than the gauge interactions
are incompatible with gauge and Lorentz symmetries [6], so that it is automatically stable thanks
to an accidental symmetry, like the proton in the SM. Other candidates that we will consider lack
this feature (i.e., some extra symmetry is responsible for their stability) but are also interesting.
The fermion 3-plet is well known from supersymmetry with matter-parity, as pure wino DM;
the fermion doublet case as pure Higgsino DM. We will also study scalar MDM candidates. In
general, this scenario has one free parameter, the DM mass M, already fixed by DM , such that
all DM signals are univocally predicted. We here address the following issues.
First, we complete the computation of the cosmological abundance performed in [6] taking into account p-wave annihilations and renormalization of the gauge couplings up to the
DM scale, and, more importantly, we include non-perturbative Sommerfeld corrections: the DM
wave-functions get distorted by Coulomb-like forces mediated by SM vectors. Their relevance
was pointed out in [7], in the case of gluino and wino as lightest supersymmetric particle. In Section 2 we discuss the basic physics of non-perturbative Sommerfeld corrections, and in Section 3
we show our results for the cosmological DM abundance.
Next, in Section 4 we compute the DM0 DM0 annihilation rates relevant for indirect MDM
signals, including the sizable enhancement due to Sommerfeld corrections.
Finally, in Section 5 we study the possibility that the ultra-high-energy cosmic rays (UHE
CR) contain some DM particle (although this looks difficult within the standard CR acceleration
mechanism), showing that MDM candidates can cross the Earth arriving to the detector in the
charged state. We discuss how it could manifest in detectors such as neutrino telescopes.
Section 6 presents our conclusions.
2. Non-perturbative Sommerfeld corrections
If the DM mass is M MZ and if the DM coupling is g g2 , DM DM annihilations into
SM vectors have a too large cross section, and consequently give a too low freeze-out DM abun M for M > MZ , the observed DM abundance is obtained for a value of M
dance. Since DM
sufficiently larger than MZ .1 Since this value turns out to be in the TeV or multi-TeV range,
we can ignore SM particle masses when computing the annihilation rates. Using SU(2) algebra,
Ref. [6] obtained a single expression for the cosmological abundance of any MDM candidate,
taking into account all (co-)annihilations in s-wave approximation. We here add p-wave annihi1 Another solution exists for M sufficiently lower than M , such that annihilation into vectors are kinematically supZ
pressed. In the minimal scenario that we consider, this solution is excluded by LEP2 and other collider data, but in general
it is still allowed in some part of the parameter range of non-minimal scenarios [8].
A DM mass M MZ can of course also be obtained by reducing g, i.e., by assuming that DM is any SU(2)L
multiplet appropriately mixed with a singlet.
154
lations and renormalization of the gauge couplings up to the DM mass M: each of these effects
gives a O(5%) correction to DM . Precise formul are given in Appendix A.
Furthermore, we take into account non-perturbative electroweak Sommerfeld corrections.
Their relevance was pointed out in [7] and might look surprising, so we start with a general semiquantitative discussion before proceeding with the detailed computations. Scatterings among
charged particles due to point-like interactions are distorted by the Coulomb force, when the kinetic energy is low enough that the electrostatic potential energy is relevant. This leads, e.g., to
significant enhancements of the + annihilation cross section (attractive force) or to significant suppressions of various nuclear processes (repulsive force). These effects can be computed
with a formalism developed by Sommerfeld [9]. In the language of Feynman graphs, these effects
are described by multi-loop photon ladder diagrams. Perturbative computations would include
only the first few diagrams: the Sommerfeld enhancement is non-perturbative because a resummation of all ladder diagrams is needed.2 The generalization of the Sommerfeld formalism to the
case of DM DM annihilations, that involves non-Abelian massive vectors, was presented in [10].
Let us first discuss Sommerfeld corrections due to one Abelian vector with mass MV and
gauge coupling : this case already contains the relevant physics and can be analyzed in a simpler
way. Cross sections at low energies are dominated by s-wave scattering, with p-wave giving
corrections of relative order T /M. Since the DM DM annihilation is local (i.e., it occurs when
the distance between the two DM particles is r 0), the non-perturbative correction is given
by R = |()/(0)|2 , where (r) is the (reduced) s-wave-function for the two-body DM DM
state with energy K, that in the non-relativistic limit satisfies the Schrdinger equation
1 d 2
+ V = K,
M dr 2
V = eMV r
r
(1)
with outgoing boundary condition ()/() iM. Here K = M 2 is the kinetic energy
of the two DM particles in the center-of-mass frame, where each DM particle has velocity .
The sign corresponds to an attractive potential and the + sign to a repulsive potential. We
define MV /M, the adimensional ratio between the vector mass and the DM mass.
Eq. (1) is the prototype of the equations that describe corrections to DM0 DM0 annihilations
mediated by a Z boson, or DM+ DM or DM+ DM+ co-annihilations mediated by a or W
or Z, or corrections mediated by a gluon (in models where coannihilations with and between
colored particles are relevant). In the three cases, the coupling that appears in Eq. (1) would
be 2 , em and s , respectively. In the non-Abelian case, Eq. (1) becomes a matrix equation and
V is the sum of the various contributions mediated by all SM vectors. Higgs exchange can also
be relevant, in models where DM sizably couples to the Higgs.
For = 0 (massless vector) the Schrdinger equation has the same form as the one that describes, e.g., the hydrogen atom, and it can be analytically solved: the Sommerfeld factor R that
multiplies the perturbative cross section is
R=
x
,
1 ex
x=
(for MV = 0).
(2)
This shows that R sizably differs from 1 at . DM DM annihilations into SM particles

freeze-out when the temperature T cools down below T /M 1/ ln(MPl /M) 1/26. This happens to be numerically comparable to the SM gauge couplings, i.e., 2 1/30. Consequently
2 Sometimes in the literature non-perturbative is used with a different meaning: to denote effects that vanish in the
perturbative limit because suppressed, e.g., by e1/ factors.
155
Fig. 1. Iso-contours of the non-perturbative Sommerfeld correction to the DM DM annihilation. Here is the coupling
constant, is the DM velocity, is the ratio between the vector mass and the DM mass. The labels indicate where
some classes of DM candidates lie in this plane: weak indicates weak-scale DM particles, TeV indicates DM with
multi-TeV mass, and strong indicates strongly-interacting particles that in some models give dominant co-annihilations.
Within the shaded region thermal masses dominate over masses, effectively shifting the value of / as indicated by the
arrow.
DM freeze-out occurs when 0.2, i.e., x 1: the Sommerfeld correction is significant,

R O(1).
Of course we must take into account that the relevant W , Z vectors are massive: since R must
now be computed numerically, it is convenient to first identify on which parameters R depends.
We notice that R only depends on the two ratios / and /, so that R can be plotted on a
plane. This can be proofed noticing that R is adimensional and that physics is invariant under
r r, M M/, MV MV /. Fig. 1 shows iso-contours of R as function of / and /.
We can distinguish various regions:
(a) Non-perturbative effects are negligible (i.e., |R 1| 1) if and/or if . We
have seen that , so non-perturbative effects are relevant when M MV /, where
MV are the SM vector masses and are their gauge couplings [10].
(b) If (lower triangle region) is so low that its value is no longer relevant:
R depends only on / showing, in the attractive case, a series of resonances. Indeed, by
increasing / the potential develops more and more bound states: resonant enhancement
happens when a bound state first appears with energy EB just below zero: in this case, at low
energy R depends on K as dictated by a BreitWigner factor 1/(K EB ).
(c) If (upper triangle region) non-perturbative effects depend almost only on /:
vector masses have a minor effect and R reduces to Eq. (2). Its value does not depend on
whether a loosely bound resonance is present or not.
When DM DM bound states are relevant, the Sommerfeld correction might not encode the full
dynamics: one should compute the density and life-time of DM DM bound states.
We include one more effect, not discussed in [7]. At finite temperature T , the , Z, W masses
and the mass splitting
M between the components of the DM multiplet are different than at zero
156
temperature. First, because these masses are proportional to the SU(2)-breaking Higgs vev v, that
depends on T and vanishes at T > Tc when SU(2) invariance gets restored by thermal effects,
via a second-order phase transition. This effect can be roughly approximated as [11]

1/2
v(T ) = v Re 1 T 2 /Tc2
(3)
.
The critical temperature Tc depends on the unknown Higgs mass (Tc mH for mH v): we
here assume Tc = 200 GeV.
Second, the squared masses of all SM vectors W , Z, , g get an extra contribution known as
thermal mass. More precisely, even in the non-Abelian case, the Coulomb force gets screened by
the thermal plasma: this can be described by a vector Debye mass, equal to [12]
11 2 2
11
(4)
m2SU(2) = g22 T 2 ,
m2SU(3) = 2g32 T 2
g T ,
6 Y
6
in the SM at T MW,Z . We approximate vector masses by summing the squared masses generated from v(T ) with the SU(2)-invariant Debye squared masses of Eq. (4).
Let us now estimate the relevance of these thermal effects. For the W and Z bosons, thermal
masses dominate at T v, and for the photon and the gluons thermal masses dominate always.
At temperature T DM particles
have a typical energy K T , so that thermal masses act like a
contribution to of order T 4 2 . When this thermal effect is relevant, in Fig. 1 one has
to shift / vertically down to the diagonal of the shaded triangle: we see that (depending on
the precise value of /) this shift has a mild or negligible effect on R, since thermal effects are
relevant when R does not have a significant dependence on .
m2U(1) =
2.1. Computing Sommerfeld corrections

Following [7,10], we now give an operative summary about how to compute Sommerfeld
corrections to s-wave DM DM annihilation rates. The set of two-body DMi DMj states that
mix among them is labeled as i, j = {1, . . . , N }. For example, when studying the wino triplet
one encounters the {DM DM+ , DM0 DM0 } system. The dynamics is encoded in the N N
potential matrix V and in another N N matrix , that describe the tree-level (co)-annihilation
rates. The strategy is the same as in the Abelian case of Eq. (1), plus a careful book-keeping of
(j )
indices. For each j one solves the following set of N coupled differential equations for i (r):
(j )
1 2 i
M r 2
N

i =1
(j )
(j )
Vii i = Ki
(5)
with boundary conditions:

(j )
(j )
i ()
= M K V ()ii i ()
(6)
r
chosen such that each wave has the same normalization at r = 0, where the annihilation occurs as
described by the matrices. The annihilation cross section i with given initial state
i at r =
(j )
i
M(KV
())r ,
is obtained by factorizing out the oscillating phase, Aij limr i (r)/e
and contracting with the annihilation matrix:

i = ci A A ii .
(7)
(j )
i (0) = ij ,
157
The factor ci is given by ci = 2 if the initial state i has two equal DM particles (e.g., as in
the DM0 DM0 state) and ci = 1 otherwise (e.g., as in the DM DM+ state). This formalism
automatically takes into account that some states cannot exist as free particles at r 1/MW ,
when the kinetic energy K is below their mass.
We next need to compute the Coulomb-like potential matrices V and the tree-level annihilation matrices for all DM components. They split into sub-systems with given values of the L
(orbital angular momentum), S (total spin) and Q (total electric charge) quantum numbers. Recalling that we neglect SM particle masses and that we consider s-wave (i.e., L = 0) tree-level
annihilations into two SM particles, annihilations obey the following selection rules:
(0) two-body DM DM states with spin S = 0 can annihilate into two SM vectors, that can have
electric charge Q = 0, 1, 2;
(1) two-body DM DM states with spin S = 1 can annihilate into two SM fermions and into two
SM Higgses, that can have electric charge Q = 0, 1.3
The initial DM DM state with S = 1 of case (1) does not exist if DM is a scalar.4
We now give explicit expressions for the annihilation matrices ii ,jj and potential matrices
Vii ,jj . Their off-diagonal entries do not have an intuitive physical meaning, and are precisely
defined in terms of the imaginary part and of the real part of the two-body propagator ii jj ,
where the indices ii denote the two DM component in the initial state, and jj denote the two
DM components in the final state.
For MDM scalars with Y = 0 one has (all states have S = 0)
ii ,jj =
Nii Njj A B A B
T , T ii T , T jj ,
8ngDM M 2
(8a)
A,B
Vii ;jj (r) = (Mi + Mj M)ii jj + Nii Njj

AB

emA r
KAB TijA TiB j + TijA TiB j
, (8b)
r
where Nij = 1 if i = j and 1/ 2 if i = j . The indices A, B run over the SM vectors

{ , Z, W + , W } with generators T A in the DM representation. The gauge couplings are included
A
in the generators, such that the gauge-covariant derivative is D = + iAA
T . We emphasize
that the (off-diagonal) entries of T are defined up to an arbitrary phase normalization for each
component of the multiplet: one needs to choose any convention such that the SU(2)-invariant
DM mass term has the same sign for all DM components. The matrices KAB and CAB respectively describe SM vector propagation and the gauge content of all other SM particles, and are
3 Non-perturbative corrections are mostly relevant when SU(2) is unbroken: in this limit one could perform a simL
plified analysis by replacing the electric charge quantum number with the SU(2)L isospin quantum number I . Then, the
only DM DM states that can annihilate into SM particles are those with S = 0 and I = 1 and 5 (that annihilate into SM
vectors; I = 1 indicates the singlet) and the state with S = 1 and I = 3 (that annihilates into SM fermions and Higgses).
4 These selection rules also apply to DM0 DM0 annihilations relevant for indirect astrophysical DM signals, and
reproduce well-known results. E.g., if DM0 is a Majorana fermion the spin-statistics relation forbids S = 1, so that
annihilations into vectors are not possible. A non-vanishing amplitude for DM0 DM0 f f, proportional to mf , is
allowed when one takes into account the small mass mf of the SM fermions f .
158
given by:
KAB
1
0
=
0
0
CAB =
25g22
4
0
1
0
0
0
0
0
1
s2
sc
0
0
0
0
,
1
0
sc 0
c 0
0 0
0 1
0
c
0 41 2 sc
+ gY
0
1
4
0
0
sc
s2
0
0
For MDM fermions with Y = 0 one has

Nii Njj A B A B
T , T ii T , T jj ,
iiS=0
,jj =
16ngDM M 2
0
0
0
0
0
0
.
0
0
(9a)
A,B
iiS=1
,jj
Nii Njj A B
=
Tii Tjj CAB ,
8ngDM M 2
(9b)
A,B
S=1 is already summed over the 3 spin components. The potentials V remain the same as in the
scalar case of Eq. (8b), up to an extra sign that must be taken into account when computing
elements of V mediated by the W , such as V0,0 = V0,0 , because an extra sign appears
in the definition of two body fermion states: |DM0 DM = |DM DM0 . Furthermore, Fermi
statistics forbids the existence of some states, such as |DM0 DM0 with S = 1.
Similar formul apply to MDM candidates with Y = 0, after taking into account that DM0
now becomes a complex particle (its normalization factor changes from N00 = 1/2 to N00 = 1)
and that one is only interested in DM DM annihilations and two body states. Summing ii ,ii
over all components one reproduces the total co-annihilation rates of [6], here reported and extended in Appendix A. Explicit expressions for V and are given in the next sections.
3. Cosmological abundances
To include non-perturbative corrections a dedicated lengthy analysis is needed for each MDM
candidate.
3.1. The fermion triplet with Y = 0 (wino)
It is not automatically stable: one needs to impose a suitable symmetry. Since Y = 0 the lightest component of this multiplet is neutral under the and under the Z, and therefore it satisfies
bounds from direct DM searches [6,13]. This multiplet appears in supersymmetric models as
SU(2)L gauginos, and behaves as MDM in limiting cases where it is much lighter than other
sparticles.
We include p-wave DM DM annihilations, but we only include non-perturbative corrections
to s-wave annihilations, that are splitted into four sectors with L = 0, total charge Q = {0, 1}
and total spin S = {0, 1}. All sectors involve one component, except the sector with Q = 0 and
S = 0, that involves two components {DM DM+ , DM0 DM0 }. In this case and in the following
we indicate the basis we employ by writing the charges of the DM components around the
and V matrices, and by indicating Q and S as pedices and apices on and V . Using Eq. (9), for
159
Fig. 2. The fermion triplet with zero hypercharge (wino). (a) (upper left): our result for its cosmological freeze-out DM
abundance as function of the DM mass. (c) (upper right) show an example (for the indicated mass M) of the temperature
dependence of the DM DM annihilation cross section, and (d) (lower right) shows the resulting cosmological evolution
of the DM abundance. (b) (lower left) shows our result for DM0 DM0 annihilation cross section relevant for indirect DM
detection, as discussed in Section 4. In each case, the continuous line is our full result, while the dashed line is the result
obtained without including non-perturbative effects.
the fermion triplet we get:

22
S=0
=
Q=0
9M 2 0
2522
,
36M 2
22
S=0
Q=1
=
,
9M 2
22
S=0
=
,
Q=2
9M 2
S=1
=
Q=0
0

3
2
,
2 2
S=0
=
VQ=0
0

2
A 2B
,
2B
0
S=1
VQ=0
= 2
A,
S=1
Q=1
=
2522
,
36M 2
S=0
VQ=2
= 2
+ A,
(10)
(11)
S=0,1
VQ=1
=
B,
(12)
(13)
2 eMZ r /r and B = eMW r /r. These results are

where
= 166 MeV, A = em /r + 2 cW
2
equivalent to those of [7].
We numerically solve the Schrdinger equations use the finite-temperature values for
and
vector masses (e.g.,
vanishes when SU(2)L -invariance is restored, etc.). Fig. 2(a) shows our
result for its freeze-out cosmological abundance: as previously noticed in [7] non-perturbative
corrections are relevant, and significantly increase the multi-TeV value of M that reproduces
the measured cosmological abundance. In the supersymmetric case, a multi-TeV wino lightest
160
supersymmetric particle implies a fine-tuning in the Higgs mass above the 103 level, so that this
scenario seems not motivated by the Higgs mass hierarchy problem.
A bound state first appears for M > M 2.5 TeV, in the DM DM system with total charge
Q = 0 and total spin S = 0. Figs. 2(c) and 2(d) show more details of the computation for M =
2.7 TeV. Since M is just above M , the bound state is loosely bound, EB 67 MeV, and gives
rise to a significant non-perturbative enhancement R of the annihilation cross section, apparent
in Fig. 2(c) as a dip at M = M . The result is reliable, because temperatures T |EB | (at which
a dedicated treatment of bound states would be necessary) do not significantly affect the final
cosmological abundance, as indicated by Fig. 2(d) and by the fact that the dip is not the main
effect.
3.2. The scalar triplet with Y = 0
This MDM candidate is not automatically stable: one needs to impose a suitable symmetry.
Having Y = 0, this candidate is compatible with bounds from direct DM searches. Any SU(2)L
multiplet of scalars can have a quartic interaction with the Higgs, H (X T a X )(H a H ), that
generates a tree-level mass splitting within the multiplet. Since this mass splitting is suppressed
by the DM mass M [6] it is not unbelievable that scalars behave as MDM because H is small
enough, H 0.05, that the
= 166 MeV mass splitting induced by gauge couplings is dominant. This is particularly plausible in the case of scalars with Y = 0, because H is not generated
by one-loop RGE corrections. Indeed RGE corrections that induce a H proportional to g22 happen to vanish, because the Higgs H is a doublet, and the doublet representation of SU(2)L has
vanishing symmetric tensor: Tr THa {THb , THc } = 0.
To include non-perturbative corrections to DM DM annihilations we split them into three
sectors, with L = 0, S = 0 and total charge Q = {0, 1, 2}. The V and matrices are
S=0
Q=0
=
222
9M 2 0
+
3
0

2
,
2
S=0
=
VQ=0
0

2
A 2B
,
2B
0
(14)
222
222
S=0
S=0
S=0
,
V
=
+
B,
=
,
VQ=2
= 2
+ A.
(15)
Q=1
Q=2
9M 2
9M 2
Fig. 3(a) shows our result for its freeze-out cosmological abundance: non-perturbative corrections are again important. Figs. 3(c) and 3(d) show details of the computation for M = 2.5 TeV,
a mass close to the critical mass M 2.5 TeV at which a bound state appears in the DM DM
system with total charge Q = 0.
We here do not study scalar and fermion triplets with |Y | = 1; the neutral component has
T3 = Y = 0 and consequently couples to the Z, giving a too large direct detection rate. A nonminimal mixing with a singlet is needed to avoid this problem. One expects significant nonperturbative corrections, similarly to the Y = 0 case, as SU(2)L gauge interactions are more
significant than U(1)Y the extra gauge interaction.
S=0
=
Q=1
3.3. The fermion quintuplet with Y = 0

This is the smallest multiplet whose lightest component is neutral under the and under
the Z and is automatically stable on the cosmological timescale. Indeed gauge couplings are
the only renormalizable interactions (compatible with gauge and Lorentz invariance) that this
161
Fig. 3. The scalar triplet with zero hypercharge. Plots have the same meaning as in Fig. 2.
multiplet can have with other SM particles. Its phenomenology is univocally dictated by a single
parameter: its Majorana mass M.
To include non-perturbative corrections to DM DM annihilations we split these scatterings
into four L = 0 sectors, with total spin S = {0, 1} and total charge Q = {0, 1, 2}. The Q = 0 and
S = 0 system is composed by 3 states:
++ 8
4A
2B
S=0
VQ=0 = + 2B
2
A
0
0
3 2B
++
3 2B ,
0
12
6
2 2

322
S=0
Q=0
=
6
9
5 2 .
2
25M
0
2 2 5 2
6
(16)
Due to Fermi statistics, the Q = 0, S = 1 system has no DM0 DM0 state:
++
S=1
VQ=0
=
+
8
4A
2B
,
2B
2
A
S=1
Q=0
=
22
4M 2
++ +
4
2

2
.
1
(17)
162
Fig. 4. The fermion quintuplet with zero hypercharge. Plots have the same meaning as in Fig. 2.
The systems with Q = 1 and S = {0, 1} have two states each:
S=0,1
VQ=1
=
S=0
=
Q=1
++
+

5
2A 6B
,
6B
3B
322
25M 2 0
++
6
+

6
,
1
(18)
S=1
=
Q=1
22
4M 2 0
++
2
+

6
.
3
(19)
Finally, in the systems with Q = 2 the tree-level annihilation rate is non-vanishing only for
S = 0:

S=0
=
VQ=2
0
+
++
4
2 3B
2 3B
,
2
+ A
S=0
=
Q=2
322 0
25M 2 +
++
+

4
12
.
12
3
(20)
Fig. 4(a) shows our result for its freeze-out cosmological abundance: the value of M that
reproduces the measured DM abundance was M 4.4 TeV. It slightly decreases down to
M 3.8 TeV after including p-wave and RGE corrections (perturbative result in Fig. 4(a)),
and increases up to almost M 10 TeV after including the non-perturbative Sommerfeld corrections. Fig. 4(d) shows that Sommerfeld corrections are mostly relevant when T
M, such
that bound states are not expected to play a relevant rle.
163
3.4. The scalar quintuplet with Y = 0

As in the scalar triplet case, is plausible to assume that the quartic coupling with the Higgs
is negligibly small, because it is not generated by RGE effects. The V and matrices are
related to those relevant for the fermion 5-plet as described in Section 2.1. Non-perturbative
corrections increase the value of the mass M that reproduces the cosmological abundance from
M 5.0 TeV [6] to M 9.4 TeV.
3.5. The scalar eptaplet with Y = 0
This is the smallest scalar multiplet that is automatically stable (its cubic scalar coupling
identically vanishes) and contains an MDM candidate compatible with all bounds. Again, it is
plausible to assume that the quartic coupling with the Higgs is negligibly small. Although we do
not show dedicated plots nor the V and matrices, we computed non-perturbative corrections
finding that they increase the value of the mass M that reproduces the cosmological abundance
from M 8.5 TeV [6] to about M 25 TeV.
3.6. The fermion doublet with Y = 1/2 (Higgsino)
It is not automatically stable, and, having Y = 0, is excluded by direct DM searches. Nevertheless, we consider it because n = 2 is the smallest multiplet, and because it is present in
supersymmetric models. The first problem can be solved by imposing a suitable symmetry, and
incompatibility with direct detection experiments can be avoided by assuming a mass mixing
with a neutral Majorana singlet (automatically present in supersymmetric models and known
as bino). We assume that this mass mixing is small enough not to affect the observables we
compute. An almost-pure Higgsino LSP is realized in corners of the MSSM parameter space: it
behaves as minimal dark matter in the sense that its properties are not dictated by supersymmetry but only by gauge invariance. In this limit, the -term is the only relevant supersymmetric
parameter, and coincides with our DM mass parameter M.
To include non-perturbative corrections to DM DM annihilations we split these scatterings
into four L = 0 sectors, with total charge Q = {0, 1} and total spin S = {0, 1}. The and V
matrices that describe each sector are
S=0
Q=0
=
22
64M 2
22
S=1
Q=0
=
128M 2 0
S=0,1
VQ=0
=
0
S=0
Q=1
=
+
3 + 2t 2
+ t4
3 2t 2 + t 4

3 2t 2 + t 4
,
3 + 2t 2 + t 4
0

+ 25
25
,
41t 4 25 41t 4 + 25
41t 4
41t 4
+
em /r
22 t 2
,
16M 2
(2c2
1)2 eMZ r /4rc2
2 eMW r /2r
S=1
Q=1
=
2522
,
64M 2

2
,
eMZ r 2 /4rc2
S=0,1
VQ=1
=+
eMW r /2r
2 eMZ r 2c2 1
,
r
4c2
(21)
(22)
164
Fig. 5. Cosmological freeze-out abundance of the fermion doublet with Y = 1/2 (Higgsino). Plots have the same
meaning as in Fig. 2, except that since non-perturbative corrections negligibly affect the cosmological abundance we do
not show details of the computations.
where t = tan W , c = cos W and

= 341 MeV is the mass splitting among charged and neutral
components generated by loop effects. Fig. 5(a) shows our result for its cosmological abundance:
in agreement with previous studies (e.g., [6]) we see that the observed DM abundance is reproduced for M = 1 TeV. Non-perturbative corrections are negligible. This also holds for the scalar
doublet with |Y | = 1/2, so we do not study it.
The Higgs potential in the minimal supersymmetric Standard Model depends on the parameter: a || = 1 TeV gives a contribution to the squared Z-mass which is 22 /MZ2 = 240 times
too large: therefore the Higgsino as minimal dark matter can only be realized at the price of a
sizable fine tuning.
4. Indirect detection of minimal dark matter
We now study the usual indirect DM signals, generated by DM0 DM0 annihilations into SM
particles. In the MDM case one only has tree-level annihilations into W + W , while the most interesting final states, and Z (both detectable as with energy practically equal to E = M),
arise at loop level. Assuming Y = 0 the cross sections (averaged over DM polarizations) are, in
the fermion DM case

2 22
DM0 DM0 W + W = n2 1
,
32M 2
2 2

2 em

2
,
DM0 DM0 = n2 1
2
16MW
(23)
and 2 times higher in the scalar DM case. These cross sections can be significantly affected by
non-perturbative Sommerfeld corrections [10], because in our galaxy DM has a non-relativistic
velocity 103 . The computation employs the same tools already described in the computation of DM DM annihilations relevant for cosmology. Non-perturbative effects can enhance the
cross section by orders of magnitude for specific values of M: the values at which the potential
develops a new bound state with Q, L, S = 0. In the triplet and quintuplet cases this enhancement
is sizable, since the DM mass suggested by cosmology is close to one of such critical values. We
show our results for the fermion triplet in Fig. 2(b), for the scalar triplet in Fig. 3(b), for the
fermion quintuplet in Fig. 4(b), and for the fermion doublet in Fig. 5(b). The vertical bands are
the 3 range of M suggested by cosmology. We have not plotted annihilation cross sections
165
Fig. 6. Spectrum of from DM0 DM0 annihilations around the galactic center, in a region of size = 103 , as seen by
a detector with E /E = 0.15. The total rate is computed assuming the NFW profile, J = 1300, and for fermion MDM
with n = 3 and M = 2.7 TeV, and for n = 5 and M = 10 TeV. The continuous line is the H.E.S.S. result [14].
into Z and ZZ since all MDM candidates with Y = 0 predict

Z = 2 / tan2 W = 6.5 ,
ZZ = / tan4 W = 10.8 .
(24)
These results allow to compute the energy spectra of annihilation products, such as photons,
antiprotons, antideutrons and positrons. Note however that, precisely for the proximity to one
of the critical values in mass, the cross sections depend strongly on the DM mass M and therefore the overall fluxes cannot be accurately predicted. Moreover, such total rates are affected by
sizable astrophysical uncertainties. For example, the number of detected photons with energy
E = M is [15]

TeV 2 2 + Z
30
J
,
N (at E = M) 2
(25)
M
m yr sr
1024 cm3 /s
where is the angular size of the observed region and the quantity J ranges between a few
and 105 , depending on the unknown DM density profile in our galaxy. Keeping these limitations
in mind, Fig. 6 shows the predicted energy spectrum of detected photons from MDM annihilations. We have assumed realistic detector parameters ( = 103 and an energy resolution of
15%). The contribution from annihilations into W + W is included using the spectral functions
computed in [10,15]. The total rate has an uncertainty of about two orders of magnitude and is
here fixed assuming the indicated value of M and the NavarroFrenkWhite DM density profile, J = 1300 [1,16], which makes this signal at the level of the present sensitivity. Indeed the
continuous line shows the spectrum of diffuse , dN /dE 5103 (TeV/E)2.3 /m2 yr TeV sr, as
observed (in a region of angular size 103 around the galactic center) by H.E.S.S. [14] up
to energies of about 10 TeV, where the number of events drops below one. Its spectral index
suggests that the observed photons are generated by astrophysical mechanisms, rather than by
DM DM annihilations. In view of the predicted value of /W W , the best MDM signal is the
peak at E = M. With the chosen parameters one gets N = 0.3 (2.6)/m2 yr for the fermion
3-plet (5-plet).
5. Minimal dark matter at ultra high energies
In this section we discuss the possibility that minimal dark matter might be detected via the
tracks left by its charged partners in experiments mainly devoted to Cosmic Ray (CR) detection,
166
such as IceCUBE [18] and km3net [17]. Indeed MDM behaves differently from other neutral
particles:
Neutrinos with energy E 1015 eV can cross the whole Earth [19] and, impinging on the
rock or ice below the detector, produce a charged partner (a muon, a tau) that cruises the
instrumented volume.
Neutral particles much heavier than neutrinos (e.g., the speculative neutralinos) interact
rarely enough that they can cross the Earth even at Ultra High Energies (UHE), E
1015 eV [20]. If they do interact before the end of their journey, they do produce a charged
partner (e.g., the speculative chargino), that however usually decays almost immediately
back to the neutral particle. For such reasons, it is generically difficult to detect them in this
way.
MDM is both heavy and quasi-degenerate with a charged partner, leading to the following behavior. To be quantitative, we focus on MDM multiplets with Y = 0. They have a stable neutral DM
particle, DM0 , and a charged DM component which is 166 MeV heavier. The latter dominantly
decays via DM DM0 with a macroscopic life-time = 44 cm/(n2 1) [6]. At ultra high
energy Lorentz dilatation makes the DM life-time long enough that a sizable fraction of the
travel is done as DM , leading to detectable charged tracks.5
This opens in principle an interesting and distinctive channel for the detection of DM. In
order to assess its feasibility in practice, we need to study the system of MDM from production to
detection. More precisely, first we study the issue of how an MDM system behaves when crossing
the Earth at UHE, having assumed that the flux of UHE CR (observed up to E 1020 GeV)
does contain some DM particles. Mechanisms to produce such UHE DM are discussed later in
Section 5.2. We finally discuss detection signatures in Section 5.3.
5.1. Propagation of UHE minimal dark matter
Consider a flux of DM0 particles that is crossing the Earth at UHE. Via Charged Current
(CC) interactions with nucleons of Earths matter, DM particles are produced. Being charged
particles traveling in a medium, these loose a part of their energy and eventually decay back to
DM0 particles. This chain of production and decay is analogous to the process that tau neutrinos
undergo in matter ( -regeneration). In the n = 5 case DM also play a role, but we will write
explicit equations for the n = 3 (wino-like) case. Such a system is described by a pair of coupled
integro-differential equations for the evolution with the position of fluxes n0 = dN0 /dE and
n = d(N+ + N )/dE of DM0 and (DM+ + DM ). They read6 :

dn0 (, E)
1
M n (E)

dCC (E , E)
dE n (E )
=
+ NN n0 (E)CC (E) +
,
d
E
2
dE
E
(26a)
5 Supersymmetric models with gravitino LSP may have a long-lived, charged next-to-lightest particle that behaves in
a way qualitatively similar to MDM [21]. Furthermore, fermionic MDM with n = 3 and Y = 0 arises in supersymmetry
in the limit of pure-wino lightest supersymmetric particle.
6 Similar equations have been of course used in the analogous problem for neutrinos [19] and for high energy neutralinos [20].
167

dn (, E)
M n (E)
1

dCC (E , E)
=
+ NN n (E)CC (E) + dE n0 (E )
d
E
2
dE
E

d
dEloss
(26b)
n
,
dE
d
where NN () is the number density profile of nucleons in the Earth [22], and E is the DM
energy. The first terms account for DM decays. The integro-differential terms within square
brackets account for DM0 DM CC scatterings. The last term accounts for the energy losses
undergone by DM , due to electromagnetic and Neutral Current (NC) scatterings:
dEloss dEloss,em dEloss,NC
=
+
.
(27)
d
d
d
We will later verify that the NC energy losses are subdominant and can be approximated as
continuous, like electromagnetic energy losses.
The cross section for the CC partonic scatterings DM0 d DM u, DM0 d DM+ u,
DM0 u DM+ d and DM0 u DM d are all given by, for fermionic MDM:

4 2
g2 (n 1) 2M 4 + 2s 2 + 2s t + t2 4M 2 s
E
d CC
d CC
,
=
=
(28)
2 )2
256
(s M 2 ) dE
d t
(s M 2 )2 (t MW
where s and t are the usual partonic Mandelstam variables. The total cross section for DM0 N
DM+ N , DM N is
1
CC (s) =
0
dx q(x)
(s M 2 )2 /s
d t
d
,
d t
(29)
where q(x) is the parton distribution function [23], summed over all quarks and anti-quarks. We
here approximated the neutron/proton fraction in Earth matter as unity. The nucleon N is at rest
with mass mN , and DM0 has energy E0 , so that s = M 2 + 2mN E0 and s = M 2 + 2xmN E0 .
A similar expression holds for NC scatterings of DM particles DM q DM q (whereas
DM0 has no NC interactions due to Y = 0):
4 2
2
d NC g2 (gLq + gRq ) 2M 4 + 2s 2 + 2s t + t2 4M 2 s
,
=
16
d t
(s M 2 )2 (t MZ2 )2
(30)
2 , g
2
where gLq = T3q Qq sW
Rq = Qq sW and q denotes any quark or anti-quark. The same
cross sections, up to terms suppressed by powers of MW,Z /M, hold for scalar MDM.
In Fig. 7(a) we plot the MDM total CC cross section. We can understand its main features as
follows. The total partonic cross section approximatively is

2
g24 (n2 1)
CC
(31)
2(xE/E ) at xE E ,
2
1
at xE E ,
128MW
where E MMW /mN 1015 eV (the transition between the two regimes happens when t
2 ). Therefore one has a constant
MW
CC 1033 cm2 at xE E . Upon integration over
the parton densities, the nucleonic CC becomes a slowly growing function of E because at
higher E partons with smaller x E /E contribute, and q(x) grows as x 0. This is analogous
168
Fig. 7. Left plot: CC cross section for fermionic MDM (black solid) with n = 3, 5 on nucleons. The same cross section
for neutrinos (green dot-dashed) is reported for comparison. Right plot: average energy loss parameters = d ln E/d
due to CC interactions (black solid lines), NC interactions (blue dashed), and electromagnetic interactions (red dotted).
For comparison, the green dot-dashed lines show the CC energy loss of , . The horizontal lines show the thickness of
the Earth, crossed vertically. (For interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
to what happens for neutrinos [24]. Indeed, at high energies, the MDM CC cross section in Fig. 7
approaches the dotted red line of the analogous result for , .
For the electromagnetic energy losses of DM one has the standard parameterization [25]
dEloss,em
(32)
= + em E,
d
where 0.24 TeV/kmwe approximates the energy loss due to ionization effects [25] and the
em term accounts for the radiative losses: it receives contributions from bremsstrahlung, e+ e
pair production and photonuclear scattering (i.e., inelastic electromagnetic collisions on nuclei).
The latter two effects have comparable importance, while bremsstrahlung is subdominant for a
very heavy particle such as MDM [26]. We follow the approach of [26] and adopt a parameterization for em with a mild dependance on the energy:

0.215
1 TeV
E
5
em 2.5 10
(33)
1+
.
kmwe M
1015 eV
Before proceeding to the numerical solution of the equations in (26), we can gain understanding on the expected results by simplifying the treatment of the CC energy losses. In analogy with
the standard parameterization, we can define an average energy loss suffered by DM0 particles
due to CC interactions as
1 dEloss,CC
= NN
CC (E) =
E
d
1
0
dx q(x)
(s M 2 )2 /s
d t
d CC
t
.
s M 2 d t
(34)
A completely analogous expression holds for NC = 1/E dEloss,NC /d. Both are plotted in
Fig. 7(b), in units of 1/kmwe = 105 cm2 /gram. The horizontal line indicates the value above
which a particle looses a significant amount of energy while vertically crossing the Earth (its
thickness being 1.1 105 kmwe). Energy losses can be understood analogously to the total cross
section, and are dominated by partons with xE E . Parton distributions have been measured
at x 104 and theoretical extrapolations to smaller x can have O(1) uncertainties, that affect
our results.
169
Fig. 8. We consider the wino-like MDM: a fermion with n = 3 and M = 2.7 TeV. Left plot: the dotted lines show the
length scales that characterize energy losses in matter with = 6.3 g/cm3 . More precisely: with the upper blue dotted
line we show the effective 1/CC experienced by a DM0 as if it never transformed in the charged state, and with the
red dotted line the range of a stable DM . The solid lines show the effects of neutral to charge transformations and vice
versa: the solid ascending red line is the DM life-time while the solid descending pink line is the mean free path for
DM0 DM CC interactions. Right plot: we consider a DM0 which crosses the Earth vertically with initial energy
Ein and we show Eout /Ein (upper red dot-dashed line) and the fraction of DM over the total number of DM particles
(DM0 + DM ). These fractions are shown as a function of Eout in the simplified numerical approach (black dashed line)
and in the full numerical approach (blue dashed line, almost superimposed to the former) as well as a function of Ein
(black solid line). (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
The fraction of energy lost in one scattering is small, /NN 102 at the UHE energies
that will be relevant for us, so that all energy losses can be approximated as continuous7 :

dE0
dE
CC
(35)
= CC E0 ,
=+
+ em + NC E ,
d
d
2
where E0 (E ) is the energy of DM0 (DM ). Therefore, we do not need to study the evolution of
the DM energy spectrum (dictated by Eq. (26)) but we just need to follow how any initial energy
decreases while crossing the Earth. This process is well approximated by the following system
of two ordinary differential equations:
1 N0
dN0
,
= NN CC N0 +
(36a)
d
1/ + 1/NN CC
dE
dE0
dE
(36b)
= N0
+ (1 N0 )
,
d
d
d
where 0 N0 1 is the fraction of DM present in neutral state, and 1 N0 is the fraction of
DM present in charged state. The initial conditions are N0 (0) = 1 and E(0) = Ein .
Fig. 8(a) shows the numerical relevance of the various effects, and allows to understand
their interplay. The DM0 interaction length starts to be smaller than the Earth thickness at
E 1015 eV, but these interactions have no effects, since the produced DM regenerates DM0
in a negligible length and with negligible energy losses. At E 1018 eV energy losses start to be
moderately relevant, and at roughly the same energy the DM0, mean free-path in Earth matter
becomes comparable to the DM decay length: the Earth is crossed loosing an order one fraction
of the initial energy, and spending an order one fraction of the path as DM rather than as DM0 .
7 The fraction of energy lost in one DM DM0 decay is
M/M 104 and can be neglected.
170
Fig. 9. We consider the MDM fermion with n = 5 and M = 10 TeV which crosses the Earth vertically with initial energy
Ein and final energy Eout . We show Eout /Ein (upper red dot-dashed line) and the fraction of DM and of DM over
the total number of DM particles as function of Eout . (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)
Fig. 8(b) shows the numerical results: the black continuous line shows the DM fraction
1 N0 as function of the initial energy, after vertically crossing the Earth. The dashed line shows
the same fraction as function of the final energy. Actually, we have two dashed lines obtained
by solving the full and simplified system of evolution equations: their agreement confirms the
validity of the continuum energy loss approximation. (In the full case we assumed an initial DM0
energy spectrum proportional to the CR energy spectrum.) This DM fraction is roughly given
by 1 N0 (1 + NN CC / )1 , and depends on the local nucleon density NN , averaged over
a typical scale of tens of km. This figure allows to compute the DM rate for any initial energy
spectrum of DM. The upper red dot-dashed line shows the ratio Eout /Ein between the final and
initial energy.
Fig. 9 shows the analogous result for the fermion 5-plet, which has a DM component with
a much shorter life-time 0.05 cm. Due to its fast decay, the new DM component is not
present with a significant abundance at the UHE CR energies E 1020 eV.
5.2. Production of UHE minimal dark matter
Minimal DM particles may exist at ultra high energies if they are a component of the primary
cosmic rays, if they manage to be produced by standard UHE CR in the Earths atmosphere or if
more exotic CR scenarios will prove to be motivated. Let us look at these opportunities in turn.
Although the mechanism that accelerates the UHE primary CR has not yet been established,
the plausible astrophysical standard mechanism (first-order Fermi acceleration in magnetic shock
waves around various objects, such as gamma-ray bursts and supernova remnants) accelerates
stable charged particles. In the MDM scenario, the neutral DM0 is accompanied by a charged
partner with life-time cm: this value is not macroscopic enough to lead to acceleration of
DM particles. Moreover, the standard mechanism accelerates protons or charged nuclei in regions that are believed to be transparent: i.e., the local density is so small these primary CR
particles negligibly hit on the surrounding DM accelerating it, or on the surrounding material
producing DM pairs.
Production of DM pairs might be relevant when UHE CR particles hit the Earth. The number
of DM particles generated by one UHE proton with energy E (2M)2 /mp 1018 eV (which is
the energy range that leads to a sizable DM fraction) is r /pp 1011 , where pp 1/m2
171
is the total hadronic cross section, and is the pp DM DM cross section, computed in [6].
A UHE electron neutrino produces a higher fraction of DM, r / 106 , but only at higher
energies E (2M)2 /me . These processes lead to a double DM signature, but their rate looks too
small for present detectors (see, e.g., [20,26]).
A UHE DM rate detectable in forthcoming experiments can arise in more speculative scenarios: e.g., they may be a component of the UHE CR generated by decays of ultra-heavy particle,
in the so-called topdown scenarios [27].
In terms of absolute numbers, if DM constitutes a number fraction r of CR with energy above
1017 eV (1018 eV), one expects a flux of 60r ( 6r) DM per km2 yr, in the case n = 3.
Atmospheric neutrinos generate a flux of 103 ( 10) up-going muons per km2 yr with E >
1013 eV (1014 eV): as discussed in the next section, this is a background to DM searches.
5.3. Detection of UHE DM
Let us study how UHE DM can be searched for. In a detector like IceCUBE or Antares DM
roughly look like muons with fake energy E = E CC / 104 E , because muon energy
losses are approximatively given by dE /d = + E with 0.2/kmwe [25]: indeed
is roughly inversely proportional to the particle mass.
One therefore needs to carefully study charged tracks to see the difference between a DM
with E 1018 eV and a muon with E 1014 eV. The muon would loose all its energy in
about 2 kmwe (with a characteristic energy loss profile) while DM have a much longer range
in matter. On the contrary, the muon is essentially stable, while DM decays in about 10 km
(less at lower energy: see Fig. 8(a)), suddenly disappearing into a DM0 and a with energy
E m E /M. Furthermore, the characteristic scale for the DM0 DM transformation is
about tens of km. Unfortunately it seems very difficult to exploit these differences to discriminate
between a and a DM in the forthcoming IceCUBE km-size detector [18], because it only
has a 1 km size and modest granularity.
6. Conclusions
We considered DM that fills one SU(2) multiplet and only interacts with SM vectors (Minimal Dark Matter [6]). This model has one free parameter, the DM mass M, which is fixed
from the observed DM abundance assuming that DM is a thermal relic. We have reconsidered
this computation including non-perturbative corrections due to Coulomb-like forces mediated by
SM vectors, that significantly increase M. For example, a fermion 5-plet automatically behaves
as MDM (and in particular is automatically stable): its mass increases from M = 4.4 TeV [6]
to 10 TeV. The cosmological freeze-out abundance is plotted as function of M in Fig. 4(a)
(fermion 5-plet), Fig. 2(a) (wino-like fermion 3-plet, M 2.7 TeV [7]), Fig. 3(a) (scalar 3-plet,
M 2.5 TeV) and Fig. 5(a) (Higgsino-like fermion 2-plet, M 1.0 TeV). We also discussed the
scalar 5-plet (M 9.4 TeV) and 7-plet (M 25 TeV).
Having fixed M, we studied the MDM signals.
Concerning the signals at colliders (see [6,29]), we here point out that the charged components DM in the MDM multiplet manifest as non-relativistic cm-scale ionizing charged tracks
negligibly bent by the B Tesla magnetic field present around the interaction region. DM decays with 97.7% branching ratio into , giving a relativistic track bent by the magnetic fields.
Everything happens in the inner portion of the detector and this signature is free from SM backgrounds. However, at a hadron collider like LHC it seems not possible to trigger on this signature:
172
Fig. 10. Predicted mass and predicted spin-independent cross sections per nucleon of MDM candidates. Scalar (fermionic) SU(2) n-tuplets are denoted as nS (nF ). The shaded region is excluded by [13] and the dashed lines indicate the
planned sensitivity of future experiments [28]. The prediction for doublets with Y = 0 hypercharge holds up to the caveats
discussed in the text; furthermore we assumed the nuclear matrix element f = 1/3 and the Higgs mass mh = 115 GeV.
level-1 triggers are located far from the interaction point and MDM candidates are so heavy that
the event rate is too low. Unlike in other scenarios we cannot choose a favorable benchmark point
in a vast parameter space.
Astrophysics offers more promising detection prospects.
The cross section for direct DM detection negligibly depends on M, and remains the same
as in [6]: Fig. 10 summarizes the situation. Since MDM is much heavier than a typical nucleus,
direct DM searches can precisely reconstruct only the combination SI /M.
Next, we considered indirect DM detection, computing how non-perturbative effects enhance
the DM0 DM0 annihilations into W + W , , Z, ZZ. Results are shown in Fig. 2(b) (winolike fermion 3-plet [10]), Fig. 3(b) (scalar 3-plet), Fig. 4(b) (5-plet), and Fig. 5(b) (Higgsino-like
fermion 2-plet [10]). Signal rates can now be computed by combining this particle physics with
(uncertain) astrophysics: for example Fig. 6 shows the predicted spectrum of galactic .
Finally, we assumed that some DM particles are present among the cosmic rays at ultra high
energies and identified an unusual signal, characteristic of heavy and quasi-degenerate multiplets
containing neutral and charged components: a DM0 crosses the Earth without loosing much
energy and, at E 1017 eV, a sizable fraction of the travel is done as DM , leading to charged
tracks, that might be detectable in neutrino telescopes such as IceCUBE or km3net [17]. Figs. 8
and 9 quantitatively show how frequent this phenomenon is.
Acknowledgements
We thank Masato Senami for several clarifications about [7]. We also thank Gianfranco
Bertone, Jrgen Brunner, Giacomo Cacciapaglia, Paschal Coyle, Michele Frigerio, Dario Grasso,
John Gunion, Teresa Montaruli, Slava Rychkov, Gnter Sigl, Igor Sokalski and Igor Tkachev for
useful discussions. The work of M.C. is supported in part by INFN under the postdoctoral grant
11067/05 and in part by CEA/Saclay.
173
Appendix A. Annihilation cross sections

We here give results for the tree-level DM DM annihilation cross sections. We define the
adimensional reduced cross section
0
(s) =
dt
s
|A |2
(A.1)
8s
where s, t are the Madelstam variables and the sum runs over all DM components and over all
SM vectors, fermions and scalars, assuming that all SM masses are negligibly small.
The DM abundance is computed by solving the Boltzmann equation
2

Y
dY
= 2 2 1 A ,
sZH z
dz
Yeq
T
A =
64 4

ds s 1/2 K1

s
A (s),
T
(A.2)
4M 2
s 1
where z = M/T , K1 is a Bessel function, Z = (1 13 gzs dg
dz ) , the entropy density of SM particles is s = 2 2 gs T 3 /45, Y = nDM /s where nDM is the number density of DM particles plus
anti-particles, and Yeq is the value that Y would have in thermal equilibrium. We can write a
single equation for the total DM density because DM scatterings with SM particles maintain
thermal equilibrium within and between the single components. We ignored the BoseEinstein
and FermiDirac factors as they are negligible at the temperature T M/26 relevant for DM
freeze-out.
We assume that DM fills one SU(2) multiplet with dimension n and hypercharge Y ; when
Y = 0 the conjugate multiplet is also added in order to allow for a gauge-invariant DM mass
term. For example, the Higgsino has n = 2 and Y = 1/2; the wino has n = 3 and Y = 0. We
define gX as the number of DM degrees of freedom: gX = n for scalar DM with Y = 0; gX = 2n
for scalar DM with Y = 0 and for spin-1/2 (Majorana) DM with Y = 0; gX = 4n for spin-1/2
(Dirac) DM with Y = 0. For fermion DM we get

gX
(9C2 21C1 ) + (11C1 5C2 ) 3
A =
24n

2 1 +
3 2C1 2 2 + C2 2 1 ln
1
4 2

4
4
2
g2 (n2 1) + 4gY4 Y 2
3g2 (n 1) + 20gY Y
3
+
+ gX
(A.3)
16
128
3
and for scalar DM we get

gX
(15C1 3C2 ) + (5C2 11C1 ) 3
A =
24n

1 +
+ 3 2 1 2C1 + C2 2 1 ln
1
4 2

4
4
2
g2 (n2 1) + 4gY4 Y 2
3g2 (n 1) + 20gY Y
+
3,
+ gX
(A.4)
48
384
where x = s/M 2 and = 1 4/x is the DM velocity in the DM DM center-of-mass frame.

The first line gives the contribution of annihilation into vectors, the second line contains the sum
174
of the contributions of annihilations into SM fermions and vectors respectively. The gauge group
factors are defined as

n(n2 1)
n(n2 1)2
Tr T A T A T B T B = gY4 nY 4 + g22 gY2 Y 2
+ g24
,
C1 =
(A.5)
2
16
A,B
C2 =

A,B
Tr T A T B T A T B = gY4 nY 4 + g22 gY2 Y 2
n(n2 1)
n(n2 1)(n2 5)
+ g24
, (A.6)
2
16
where the sum is over all SM vectors A = {Y, W 1 , W 2 , W 3 } with gauge coupling generators T A .
The DM freeze-out abundance is accurately determined by the leading two terms of the expansion of for small , that describe the s-wave and the p-wave contributions. This approximation
allows to analytically do the thermal average in Eq. (A.2):
0
A cs + cp 3 + implies

0 MT 3 e2M/T
3T
cs
+
+
A
c
+
.
c
s
p
2M
2
32 3
(A.7)
More complex models of DM can have extra interactions beyond gauge interactions, but the
gauge contribution we computed accurately is the minimal model-independent contribution
present in all cases. Eq. (A.2) also arises as a subset of the Boltzmann equations that describe
leptogenesis from decays of gauge-interacting particles: the scalar triplet corresponds to n = 3,
Y = 1 [30], and the fermion triplet to n = 3 and Y = 0 [31].
References
[1] See, G. Bertone, D. Hooper, J. Silk, Phys. Rep. 405 (2005) 279, hep-ph/0404175 for a recent review.
[2] These numbers summarize various recent global analyses of cosmological data within the CDM model that found
compatible values and uncertainties: D.N. Spergel, et al., WMAP Collaboration, astro-ph/0603449;
M. Cirelli, A. Strumia, JCAP 0612 (2006) 013, astro-ph/0607086;
M. Tegmark, et al., Phys. Rev. D 74 (2006) 123507, astro-ph/0608632.
[3] A. Gould, et al., Phys. Lett. B 238 (1990) 337.
[4] Strongly interacting, hadronized, Dark Matter is subject to a number of constraints analysed in G.D. Starkman,
A. Gould, R. Esmailzadeh, S. Dimopoulos, Phys. Rev. D 41 (1990) 3594, that leave only implausible windows open;
Recently, G.D. Mack, J.F. Beacom, G. Bertone, 0705.4298 went in the direction of closing those windows to safeguard Earths heat flow.
[5] The upper bound is imposed by the null results of direct detection searches, see [13]. See also G. Belanger,
A. Pukhov, G. Servant, 0706.0526 for a recent analysis of this feature as a phenomenological assumption and
in specific models.
[6] M. Cirelli, N. Fornengo, A. Strumia, Nucl. Phys. B 753 (2006) 178, hep-ph/0512090.
[7] The relevance of non-perturbative electroweak corrections to DM freeze-out was pointed out in J. Hisano, S. Matsumoto, M. Nagai, O. Saito, M. Senami, Phys. Lett. B 646 (2007) 34, hep-ph/0610249;
Non-perturbative QCD corrections were considered in H. Baer, K. Cheung, J. Gunion, Phys. Rev. D 59 (1999)
075002, hep-ph/9806361.
[8] R. Barbieri, L. Hall, V. Rychkov, Phys. Rev. D 74 (2006) 015007, hep-ph/0603188;
A. Pierce, J. Thaler, hep-ph/0703056;
F. DEramo, 0705.4493;
L. Lopez Honorez, E. Nezri, J.F. Oliver, M.H.G. Tytgat, JCAP 0702 (2007) 028, hep-ph/0612275;
R. Enberg, P. Fox, L. Hall, A. Papaioannou, M. Papucci, 0706.0918.
[9] A. Sommerfeld, Ann. Phys. 11 (1931) 257.
[10] J. Hisano, S. Matsumoto, M.M. Nojiri, O. Saito, Phys. Rev. D 71 (2005) 015007, hep-ph/0407168;
J. Hisano, S. Matsumoto, M.M. Nojiri, Phys. Rev. D 67 (2003) 075014, hep-ph/0212022;
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
175
J. Hisano, S. Matsumoto, M.M. Nojiri, Phys. Rev. Lett. 92 (2004) 031303, hep-ph/0307216;
J. Hisano, S. Matsumoto, M.M. Nojiri, O. Saito, Phys. Rev. D 71 (2005) 063528, hep-ph/0412403;
J. Hisano, S. Matsumoto, O. Saito, M. Senami, Phys. Rev. D 73 (2006) 055004, hep-ph/0511118.
An approximated discussion of the SM potential at finite temperature can be found in M. Dine, R. Leigh, P. Huet,
A. Linde, D. Linde, Phys. Rev. D 46 (1992) 550, hep-ph/9203203.
D.J. Gross, R.D. Pisarski, L.G. Yaffe, Rev. Mod. Phys. 53 (1981) 43;
H. Weldon, Phys. Lett. D 26 (1982) 1394.
Xenon-10 Collaboration, 0706.00393.
H.E.S.S. Collaboration, Nature 439 (2006) 695, astro-ph/0603021;
See L. Bergstrom, T. Bringmann, M. Eriksson, M. Gustafsson, AIP Conf. Proc. 861 (2006) 814, astro-ph/0609510
for a recent summary of -line signals of some DM candidates.
L. Bergstrom, P. Ullio, J. Buckley, Astropart. Phys. 9 (1998) 137, astro-ph/9712318.
J. Navarro, C. Frenk, S. White, Astrophys. J. 462 (1996) 563, astro-ph/9508025.
See http://km3net.org and http://antares.in2p3.fr.
IceCube Collaboration, AIP Conf. Proc. 842 (2006) 971, astro-ph/0601269.
S. Iyer, M.H. Reno, I. Sarcevic, Phys. Rev. D 61 (2000) 053003, hep-ph/9909393;
S.I. Dutta, M.H. Reno, I. Sarcevic, Phys. Rev. D 62 (2000) 123001, hep-ph/0005310;
A. LAbbate, T. Montaruli, I. Sokalski, Astropart. Phys. 23 (2005) 57, hep-ph/0406133;
For a summary of the formalism, including the effect of oscillations relevant at sub-TeV energies, see M. Cirelli, N.
Fornengo, T. Montaruli, I. Sokalski, A. Strumia, F. Vissani, Nucl. Phys. B 727 (2005) 99, hep-ph/0506298.
S. Bornhauser, M. Drees, Astropart. Phys. 27 (2007) 30, hep-ph/0603162.
I. Albuquerque, G. Burdman, Z. Chacko, Phys. Rev. Lett. 92 (2004) 221802, hep-ph/0312197;
M. Ahlers, J. Kersten, A. Ringwald, JCAP 0607 (2006) 005, hep-ph/0604188;
I.F.M. Albuquerque, G. Burdman, Z. Chacko, Phys. Rev. D 75 (2007) 035006, hep-ph/0605120;
M. Ahlers, J.I. Illana, M. Masip, D. Meloni, 0705.3782.
A.M. Dziewonski, D.L. Anderson, Phys. Earth Planet. Interior 25 (1981) 297;
See also http://mahi.ucsd.edu/Gabi/rem.html.
We use the parton distribution functions from S. Alekhin, JETP Lett. 82 (2005) 628, hep-ph/0508248, that are
available in Mathematica format from http://sirius.ihep.su/alekhin/pdfa02 and extend down to small x 107 , as
needed for the computation of nucleonic cross sections at UHE.
For a review and references, see Section 4.3 of A. Strumia, F. Vissani, hep-ph/0606054.
P.H. Barrett, L.M. Bollinger, G. Cocconi, Y. Eisenberg, K. Greisen, Rev. Mod. Phys. 24 (1952) 133;
D.E. Groom, N.V. Mokhov, S.I. Striganov, At. Data Nucl. Tables 78 (2001) 183.
M.H. Reno, I. Sarcevic, S. Su, Astropart. Phys. 24 (2005) 107, hep-ph/0503030.
V. Berezinsky, M. Kachelriess, A. Vilenkin, Phys. Rev. Lett. 79 (1997) 4302, astro-ph/9708217.
For the Xenon project see E. Aprile, et al., Nucl. Phys. B (Proc. Suppl.) 138 (2005) 156, astro-ph/0407575;
For the SuperCDMS project see CDMS-II Collaboration, eConf C041213 (2004) 2529, astro-ph/0503583. Useful
comparisons can be done using the tools in http://dendera.berkeley.edu/plotter/entryform.html.
J. Gunion, S. Mrenna, Phys. Rev. D 62 (2000) 015002, hep-ph/9906270.
T. Hambye, M. Raidal, A. Strumia, Phys. Lett. B 632 (2006) 667, hep-ph/0510008.
T. Hambye, Y. Lin, A. Notari, M. Papucci, A. Strumia, Nucl. Phys. B 695 (2004) 169, hep-ph/0312203. Their cross
section into gauge bosons should be increased by a factor 2, and this has a minor effect on the baryon asymmetry.
The common origin of linear and nonlinear chiral

multiplets in N = 4 mechanics
F. Delduc a, , E. Ivanov b
a Laboratoire de Physique de lENS Lyon, CNRS UMR 5672, 46 Alle dItalie, 69364 Lyon Cedex 07, France
b Bogoliubov Laboratory of Theoretical Physics, JINR, 141980 Dubna, Moscow region, Russia

Abstract
Elaborating on previous work [F. Delduc, E. Ivanov, Nucl. Phys. B 753 (2006) 211, hep-th/0605211;
F. Delduc, E. Ivanov, Nucl. Phys. B 770 (2007) 179, hep-th/0611247], we show how the linear and nonlinear
chiral multiplets of N = 4 supersymmetric mechanics with the off-shell content (2, 4, 2) can be obtained
by gauging three distinct two-parameter isometries of the root (4, 4, 0) multiplet actions. In particular,
two different gauge groups, one Abelian and one non-Abelian, lead, albeit in a disguised form in the second
case, to the same (unique) nonlinear chiral multiplet. This provides an evidence that no other nonlinear
chiral N = 4 multiplets exist. General sigma model type actions are discussed, together with the restricted
potential terms coming from the FayetIliopoulos terms associated with Abelian gauge superfields. As in
our previous work, we use the manifestly supersymmetric language of N = 4, d = 1 harmonic superspace.
A novel point is the necessity to use in parallel the and gauge frames, with the bridges between these
two frames playing a crucial role. It is the N = 4 harmonic analyticity which, though being non-manifest
in the frame, gives rise to both linear and nonlinear chirality constraints.
PACS: 11.30.Pb; 11.15.-q; 11.10.Kk; 03.65.-w
Keywords: Supersymmetry; Gauging; Isometry; Superfield
E-mail addresses: francois.delduc@ens-lyon.fr (F. Delduc), eivanov@theor.jinr.ru (E. Ivanov).

doi:10.1016/j.nuclphysb.2007.07.015
F. Delduc, E. Ivanov / Nuclear Physics B 787 (2007) 176197
177
1. Introduction
The one-dimensional supersymmetry and related models of supersymmetric (quantum) mechanics reveal many specific surprising features and, at the same time, have a lot of links with
higher-dimensional theories of current interest (see, e.g., [1,2] and references therein). This motivates many research groups towards thorough study and further advancing of this subject (see,
e.g., [2,3]).
Recently, we argued [4,5] that the plethora of relationships between various d = 1 supermultiplets with the same number of fermionic fields, but different divisions of the bosonic fields
into physical and auxiliary ones (so-called d = 1 automorphic dualities [6]), can be adequately
understood in the approach based on the gauging of isometries of the invariant actions of some
basic (root) multiplets by non-propagating (topological) gauge multiplets (these isometries
should commute with supersymmetry, i.e., be triholomorphic). The key merit of our approach
is the possibility to study these relationships in a manifestly supersymmetric superfield manner,
including the choice of supersymmetry-preserving gauges. Previous analysis [6,7] was basically
limited to the component level and used some ad-hoc substitutions of the auxiliary fields. In
the framework of the gauging procedure, doing this way corresponds to making use of the Wess
Zumino-type gauges.
In [4,5] we focused on the case of N = 4 supersymmetric mechanics and showed that the
actions of the N = 4 multiplets with the off-shell contents (3, 4, 1), (1, 4, 3) and (0, 4, 4) can be
obtained by gauging certain isometries of the general actions of the root multiplet (4, 4, 0) in
the N = 4, d = 1 harmonic superspace. Based on this, we argued that the latter is the underlying
superspace for all N = 4 mechanics models. Here we confirm this by studying, along the same
line, the remaining N = 4 multiplets, the chiral and nonlinear chiral off-shell multiplets with the
field content (2, 4, 2). They prove to naturally arise as a result of gauging some two-parameter
isometry groups admitting a realization on the harmonic analytic superfield q +a which describes
the multiplet (4, 4, 0). The origin of the difference between these two versions of the (2, 4, 2)
multiplet is attributed to the fact that they emerge from gauging two essentially different isometries: the linear multiplet is associated with gauging of some purely shift isometries, while the
nonlinear one corresponds to gauging the product of two rotational isometries, viz. the target
space rescalings and a U (1) subgroup of the SU(2) PauliGrsey group. We also recover the
same nonlinear version of this multiplet, though in disguise, by gauging a mixture of the rescaling and shift isometries. It is the last possible two-parameter symmetry implementable on q +a .
Thus we show that two known off-shell forms of the N = 4, d = 1 multiplet (2, 4, 2), the linear
and nonlinear ones, in fact exhaust all possibilities.
It should be emphasized that the existence of nonlinear cousins of the basic d = 1 supermultiplets is one of the most amazing features of extended d = 1 supersymmetry. In superspace
they are described by superfields satisfying some nonlinear versions of the standard constraints
(e.g. of chirality constraints). As a result, the realization of the corresponding off-shell d = 1
supersymmetry on the component fields of such supermultiplets is intrinsically nonlinear. The
list of such multiplets known to date includes the nonlinear analogs of the N = 4 multiplets
(4, 4, 0) [4,811], (3, 4, 1) [12,13], (2, 4, 2) [13], as well as of the N = 8 multiplets (4, 8, 4)
[14] and (2, 8, 6) [15]. As shown in [4], within the approach based on the gauging procedure the
difference between linear and nonlinear (3, 4, 1) multiplets originates from the fact that the first
multiplet is related to the gauging of the shift or rotational U (1) symmetries, and the second one
to the gauging of the target space rescalings. In the case of the (2, 4, 2) multiplets we find an
178
analogous intimate relation between the type of multiplet and the q +a two-parameter symmetry
group which has to be gauged to generate it.
2. Brief preliminaries
Throughout the paper we use the same N = 4, d = 1 harmonic superspace (HSS) techniques,
conventions and notation as in [4,5,12]. The N = 4 root multiplet (4, 4, 0) is described by
the harmonic analytic superfield q +a (, u) which is subjected to the Grassmann harmonic and
bosonic harmonic constraints
(a)
D + q +a = D + q +a = 0,
(b)
D ++ q +a = 0.
(2.1)
Here (, u) are coordinates of the harmonic analytic N = 4 superspace [12], (, u) =

+i
(tA , + , + , u
i ), u ui = 1, they are related to the standard N = 4 superspace (central basis)
coordinates z = (t, i , i ) as

i
tA = t i + + + , = i u
(2.2)
i , = ui .
Respectively, the N = 4 covariant spinor derivatives and their harmonic projections are defined
by
+ i i t ,
D i = i + ii t ,
i

i
D i = D i ,
D , D k = 2i i t ,
Di =
(2.3)
i
D = u
i D ,
i
D = u
i D ,

+
= D , D + = 2itA .
D , D
(2.4)
In the analytic basis, the derivatives D + and D + are short,

D+ =
D + = ,

(2.5)
so the conditions (2.1)(a) become the harmonic Grassmann CauchyRiemann conditions stating
that q +a does not depend on the coordinates , in this basis. The analyticity-preserving
harmonic derivative D ++ and its conjugate D in the analytic basis are given by
+ + ,

D = 2i tA + + + + ,
D ++ = ++ 2i + + tA + +
= u
i
,
u
i
(2.6)
and are reduced to the pure harmonic partial derivatives in the central basis. They satisfy
the commutation relations
++
0
D ,D
(2.7)
= D0,
= 2D ,
D ,D
where D 0 is the operator counting external harmonic U (1) charges. In the analytic basis it is
given by
D 0 = u+
i

+
+
,
+ ui
+ + +
+

ui
ui
(2.8)
179
while in the central basis it coincides with its pure harmonic part. On the extra doublet index a of
the superfield q +a the so-called PauliGrsey group SU(2)PG is realized. It commutes with the
N = 4 supersymmetry generators, as distinct from the R-symmetry SU(2)R group which acts
on the doublet indices i, k of the Grassmann and harmonic coordinates, spinor derivatives and
N = 4 supercharges.
The free action of q +a can be written either in the analytic, or the central superspace

i
+a +
1
H q +a qa =
(2)
q t qa ,
Sqfree =
(2.9)
A
4
2
where q a D q +a and the integration measures are defined as

+ +
H = du dt d 4 = du dtA D D D + D + = (2)
D D ,
A

(2)
A = du d (2) = du dtA d + d + = du dtA D D .
The general sigma model-type action of
given by

Sq = H L q +a , q b , u .
q +a
(2.10)
(with a non-trivial bosonic target space metric) is
(2.11)
The constraint (2.1) possesses a seven-parameter group of rigid symmetries commuting with
supersymmetry (it includes SU(2)PG as a subgroup) [5]. One can single out the appropriate subclasses of the general action (2.11) (including the free action (2.9)) which are invariant with
respect to one or another symmetry of this sort. For further use we give here the full list of
non-equivalent two-parameter symmetries.
Abelian symmetries
(a)
q +a = 1 u+a + 2 c(ab) u+
b,
(b) q +a = 1 q +a 2 c(ab) qb+ .
(2.12)
Non-Abelian symmetry
q +a = 1 q +a + 2 u+a .
Here the constant triplet
c2 = cab cab = 2.
cab
is normalized
(2.13)
as1
(2.14)
The algebra of the transformations (2.13) provides an example of two-generator solvable algebra.
All other possible two-parameter symmetry groups listed in [5] can be reduced to (2.12), (2.13)
by a redefinition of q +a .
In what follows we shall gauge these symmetries and show that this gauging gives rise to three
versions of the N = 4 multiplet (2, 4, 2), with the corresponding general actions arising from the
appropriate invariant subclasses of the general q +a action (2.11). It turns out that the standard
linear chiral N = 4 multiplet emerges as the result of gauging purely shift isometry (2.12)(a)
while the two alternative gaugings give rise to two nonlinear versions of this multiplet. The
nonlinear multiplet obtained from gauging the group (2.12)(b) is identical to the one discovered
in [13]. The multiplet obtained from gauging (2.13), although looking different, can be identified
with the previous one after suitable redefinitions.
1 We use the same notation for the unrelated constant triplets in (2.12)(a) and (2.12)(b), hoping that this will not give
rise to any confusion.
180
3. Chiral multiplet
Our gauging prescriptions are basically the same as in other cases [4,5].
We start with the gauged version of the transformations (2.12)(a)
q +a = 1 u+a + 2 cab u+
b,
cab = cba ,
(3.1)
where 1 and 2 are now charge-zero analytic superfields. The gauge covariantization of the
q +a harmonic constraint (2.1)(b) is given by
D ++ q +a V1++ u+a V2++ cab u+
b = 0,
(3.2)
where V1++ and V2++ are analytic gauge superfield transforming as

V1++ = D ++ 1 ,
V2++ = D ++ 2 .
(3.3)
In the cases considered here (as distinct from the cases treated in [4,5]), it is convenient to
make use of the bridge representation of the gauge superfields [16,17].
The analytic superfields V1++ and V2++ may be represented as
V1++ = D ++ v1 ,
V2++ = D ++ v2 ,
(3.4)
where v1 (t, , , u) and v2 (t, , , u) are non-analytic harmonic superfields which may be interpreted as bridges between the analytic and central superspace gauge groups (called the and
gauge groups, see [17]). They transform under local shifts as
v1 = 1 (, u) 1 (t, , ),
v2 = 2 (, u) 2 (t, , ),
(3.5)
where 1 (t, , ) and 2 (t, , ) are non-analytic gauge superparameters bearing no dependence
on the harmonic variables
D ++ 1 (t, , ) = 0,
D ++ 2 (t, , ) = 0.
(3.6)
We now define the non-analytic doublet superfield Q+a by

Q+a = q +a v1 u+a v2 cab u+
b.
(3.7)
As a consequence of (3.2), (3.4), it satisfies the simple harmonic constraint

D ++ Q+a = 0,
(3.8)
which implies
Q+a (t, , , u) = Qba (t, , )u+
b,
where the superfields
quartet. We also have
Qba (t, , )
Qba = Qba ,
(3.9)
are independent of the harmonic variables and form a real
a
v1 ua v2 cab u
Qa D Q+a = Qba (t, , )u
b =q
b,
(3.10)
where
q a D q +a V1 ua V2 cab u
b
and
(3.11)
V1 = D v1 ,
181
V2 = D v2 ,
V1 = D 1 ,
V2 = D 2 .
(3.12)
From (3.10) and (3.12) one can find the gauge transformation law of the non-analytic harmonic
superfield q a :
q a = 1 ua + 2 cab u
b.
(3.13)
Qa
introduced in (3.7), (3.10) transform unNow it is easy to determine how the superfields
der the local shift symmetries. They are inert under the gauge transformations and have the
following gauge transformation law
Qa = 1 ua + 2 cab u
b
Qba = 2 cab 1 ba .
(3.14)
As a consequence of the Grassmann analyticity constraints (2.1), the superfield Q+a satisfies
the following fermionic constraints

D + q +a = 0 D + Q+a + D + v1 u+a + D + v2 cab u+
b = 0,
+ +a + ab +
+
+a
+
+a
D q = 0 D Q + D v1 u + D v2 c ub = 0.
(3.15)
It is important that, due to the analyticity of the gauge superfields VI++ , I = 1, 2, the fermionic
connections D + vI , D + vI depend linearly on the harmonic variables. We shall use the notations
D + vI (t, , u) = AaI (t, , )u+
a,
D + vI (t, , u) = A aI (t, , )u+

a.
(3.16)
Using (3.9) and (3.16), one can rewrite (3.15) in the following equivalent form with no harmonic
dependence at all:
(a
(a
D (a Qb)c A1 b)c + A2 cb)c = 0,
(a
(a
D (a Qb)c + A 1 b)c A 2 cb)c = 0.
(3.17)
Now let us more closely inspect the transformation laws (3.14). We start by choosing a frame
in which the matrix cab has only one non-vanishing component:
c12 = c21 = i,
c11 = c22 = 0.
(3.18)
In this frame the transformations (3.14) look as

Q1 = (1 i2 )u1 ,
Q2 = (1 + i2 )u2 ,
or, in terms of the N = 4 superfields

Q12 = (1 + i2 ),
Qab (t, , )
(3.19)
defined in (3.9),
Q21 = (1 i2 ),
Q11 = Q22 = 0.
(3.20)
It is then convenient to choose the unitary-type gauge

Q12 = Q21 = 0.
(3.21)
AaI
and their conjugate

Now the constraints (3.17) determine the spinor connection superfields
a
11
AI in terms of the remaining superfield Q and its complex conjugate Q22 =

1
A11 = D 2 ,
2
i
1
A2 = D 2 ,
2
A21 = D 1 ,
2
i
A22 = D 1 (and c.c.),
2
(3.22)
182
simultaneously with imposing the constraints on these superfields

D 1 = D 1 = 0,
D 2 = D 2 = 0.
(3.23)
These constraints may be interpreted as twisted chirality conditions. They can be given the standard form of the chirality conditions by relabelling the spinor derivative D i , D i in such a way that
the R-symmetry SU(2) acting on the indices i gets hidden, while another SU(2) (which rotates
D i through D i ), gets manifest.2 Thus we have succeeded in deriving the linear chiral N = 4,
d = 1 multiplet (2, 4, 2) from the analytic multiplet (4, 4, 0) by gauging two independent shift
isometries realized on the latter.
Let us now examine this correspondence on the level of the invariant actions. We start with
the gauge covariantization of the free action (2.9) of the analytic superfield q +a . In the full
superspace the covariantized action reads

+a b +
free
Scov
= H q +a D qa+ 2V1 q +a u+
a 2V2 q ca ub

+ 2 V1++ V2 V2++ V1 c+ .
(3.24)
It is gauge invariant up to a total harmonic derivative in the integrand, i.e., it is of the Chern
Simons type. It may be equivalently rewritten in terms of the superfield Q+a and the bridges
v1 , v 2

free
+a
+a b
Scov
= H Q+a D Q+
a + 2v1 Q ua + 2v2 Q ca ub

+ v1 V2++ v2 V1++ c + v12 + v22 .
(3.25)
In this form it is invariant under both the and gauge transformations. In fact, the action can
be written as a sum of two terms, the first of which transforms only under the gauge group, and
the second only under the group
free
Scov
= S + S ,

2
1 +a
S = H Q+a D Q+
Q ua Qa u+
a
a
4

1
a b + 2
Q+a ca b u
Q
c
u
,
a b
b
4

2 1 +a b
2
1 +a
S = H
+ q ca ub q a ca b u+
q ua q a u+
a
b
4
4
++ a b
++ a
V2 q ca ub V1 q ua .
To check the invariance of S we use the following transformation laws:

a +
a b +
Q+a u
Q+a ca b u
a Q ua = 21 ,
b Q ca ub = 22 ,

+a

+a b

a +
a b +
Q+a Q
a = 1 Q ua Q ua + 2 Q ca ub Q ca ub .
(3.26)
(3.27)
(3.28)
2 Both these R-symmetry SU(2) are manifest in the quartet notation D ii = (D i , D

i ) = (D i , D i ), D ii = Dii =
(D i , Di ) = (D i , D i ).
183
It is worthwhile to note that all three terms in the action

S are SU(2) singlets (independent

of harmonic variables), so that the harmonic integral du is in fact not necessary. After some
simple algebra, making use of the integration by parts with respect to the harmonic derivatives,
the constraint (3.2) and the definitions (3.4), (3.11) and (3.12), one can show that, up to a total
harmonic derivative,
1 free
S = Scov
(3.29)
,
2
whence one obtains the representation of the gauge-covariantized q + action (3.24) solely in terms
of the superfields Qa :

2
1 +a
free
Scov
= 2S = 2 H Q+a Q
Q ua Qa u+
a
a
4
2
1 +a b
Q ca ub Qa ca b u+
(3.30)
.
b
4
Now, the gauge condition (3.21) simply amounts to
a +
Q+a u
a Q ua = 0,
a b +
Q+a ca b u
b Q ca ub = 0,
(3.31)
and in this gauge the action (3.30) takes the standard form of the free action of the (twisted)
chiral (2, 4, 2) multiplet

free
Scov
(3.32)
= 2 H Q+a Q
=
2
dt d 4 .
a
Thus the free action of the chiral N = 4 multiplet arises as a particular gauge of the properly
gauge-covariantized free action of the analytic multiplet q +a . Note that this equivalence, like in
other cases [4,5], was shown here in a manifestly N = 4 supersymmetric superfield approach,
without any need to pass to the components. It is interesting to note that there exist two more
equivalent useful forms of the action (3.24) in terms of the original superfields q a :

++ a b
free
q ca ub
Scov
= H q +a qa V1++ q a u
a V2

= H D ++ q a qa .
(3.33)
Checking the gauge invariance of the action in the second form is especially simple: one uses the
transformation law (3.13) and the fact that D ++ q a is analytic in virtue of the relation
D ++ q a = q +a + V1++ ua + V2++ cab u
b.
(3.34)
Let us now comment on the general sigma-model type action. The only invariant of the
gauge transformations (3.1), (3.13) which one can construct from q +a and q a is the quantity X
defined as follows

a ++
X = q +a u+
q ua c ,
a c

a + +a
D X = 2 q ua c q ua + q a u+
,
a c

2
2
D ++ D X = 2X,
(3.35)
D ++ X = D X = 0,
where
c = cab u
a ub ,
etc.
184
It admits the equivalent representation, in which its gauge invariance becomes manifest
a ++
Q ua c = 2Q(ad) cdb u+
X = Q+a u+
(3.36)
a c
(a ub)
(the remaining relations in (3.35) preserve their form modulo the replacements q a Qa ).
The subclass of the general sigma-model type q + action (2.11) invariant under the gauge transformations (3.1), (3.13) is then defined as follows

Scov = H L X, D ++ X, D X, u .
(3.37)
In the gauge (3.21) and in the frame (3.18) we have

+
X = 2i u+
1 u1 u2 u2
(3.38)
and, after integration over harmonics, the action (3.37) is reduced to the most general action of
Note the relation
the twisted chiral superfields , .

1 ++ 1 + +
+
u
Xu
u
Xu
u
D
D
= i Xu
(3.39)
2 2
2 2
2 2 .
2
2
The free action (3.24) can also be expressed through the universal invariant X:

3
free
H X 2 .
=
Scov
(3.40)
2
This relation can be proved, starting from the form of the free action (3.30) and integrating by
parts with respect to the harmonic derivatives. Another way to see this is to compare both sides in
the original frame representation (i.e., in terms of q a ) by choosing the WessZumino gauge
for the superfields V1 and V2
VI = BI .
Finally, we address the issue of the FayetIliopoulos (FI) terms. In the present case one can
define two independent gauge invariant FI terms

S1FI = i1 du d (2) V1++ ,
(3.41)
S2FI = i2 du d (2) V2++ .
Let us consider the first one. Rewriting it in the full harmonic superspace

FI
++
S1 = i1 H D v1 = i1 H + + + v1 ,
(3.42)
expressing v1 from the relation (3.7), integrating by parts with using the analyticity property of
q +a and V1++ , V2++ and performing in the end the integration over harmonics, this term in the
gauge (3.21) and frame (3.18) can be transformed to the expression

i
S1FI = 1 dt d 4 1 2 2 1 .
(3.43)
2
It is a particular case of twisted chiral superpotential term. Analogously,

1
S2FI = 2 dt d 4 1 2 + 2 1 .
(3.44)
2
It is unclear whether a general chiral superpotential can be generated from some gauge invariant
q +a action.
185
4. Nonlinear chiral multiplet

Let us now consider the gauging of the two-parameter Abelian symmetry (2.12)(b)
q +a = 1 q +a + 2 cba q +b ,
I = I (, u),
I = 1, 2.
(4.1)
The harmonic constraint (2.1)(b) is now covariantized as

D ++ q +a V1++ q +a V2++ cba q +b = 0.
(4.2)
The analytic potentials VI++ , I = 1, 2, possess the same gauge transformation laws (3.3) and
are expressed through the bridges vI with the mixed transformation rules (3.5) by the same
relations (3.4). However, since now q +a transforms homogeneously under the gauge transformations, the relation (3.7) between the and world objects has to be modified:

q +a = ev1 cos v2 Q+a + sin v2 cba Q+b ,

Q+a = ev1 cos v2 q +a sin v2 cba q +b ,
(4.3)
or, in another form,

q +a + icba q +b = ev1 iv2 Q+a + icba Q+b ,

q +a icba q +b = ev1 +iv2 Q+a icba Q+b .
(4.4)
A direct calculation shows that the constraint (4.2) entails, for Q+a ,
D ++ Q+a = 0,
(4.5)
whence, in the central basis,

Q+a = Qia (t, , )u+
i
(4.6)
(cf. (3.9)). Also, using the transformation laws (3.5) and (4.1), it is easy to find
Qa = 1 Qa + 2 cba Qb .
(4.7)
In what follows it will be convenient to choose the SU(2) frame (3.18) in which
q +1 = e(v1 +iv2 ) Q+1 ,
q +2 = e(v1 iv2 ) Q+2 ,
Q11 = (1 + i2 )Q11 ,
Q22 = (1 i2 )Q22 ,
Q21 = (1 + i2 )Q21 ,
Q12 = (1 i2 )Q12 .
Like in Section 3, the analyticity of

D + q +1 = D + q +1 = 0
D + q +2 = D + q +2 = 0
q +a
implies the covariant analyticity for

D + + D + (v1 + iv2 ) Q+1

= D + + D + (v1 + iv2 ) Q+1 = 0,
+

D + D + (v1 iv2 ) Q+2

= D + + D + (v1 iv2 ) Q+2 = 0.
(4.8)
(4.9)
Q+a :
(4.10)
(4.11)
++
it follows that the gauge connections in (4.10), (4.11) are linear in
From the analyticity of V1,2
harmonics
186
D + (v1 + iv2 ) = Aa(+) (t, , )u+

a,
D + (v1 iv2 ) = Aa() (t, , )u+

a,
D + (v1 iv2 ) = A a(+) (t, , )u+

a,
D + (v1 + iv2 ) = A a() (t, , )u+

a.
(4.12)
Now it is time to properly fix the gauge freedom (4.9). Assuming that Q12 = (Q21 ) possesses
a non-zero constant background, a convenient gauge is
Q12 = Q21 = 1.
(4.13)
Substituting this gauge into the covariant analyticity conditions for Q+a in (4.11), taking into
account the relations (4.6), (4.12), and equating to zero the coefficients of three independent
+ 2
+ +
2
products of harmonics ((u+
1 ) , (u2 ) and u1 u2 ), we obtain
A2(+) = A1() = 0,
A2() = D 1
A1(+) = D 2 ,
(and c.c.),
D + D = D 1 + D 2 = 0,
1 = D 2 D 1 = 0,
D 2 D
1
(4.14)
where Q11 , Q22 = .

Thus the only independent object that remains in the gauge (4.13) is a complex N = 4 superfield subjected to the constraints (4.14). These constraints are a nonlinear version of the
twisted chirality constraints (3.23) and are easily recognized as a twisted version of the nonlinear N = 4 chirality constraints [13]. It can be given the form of the ordinary nonlinear chirality
constraints by relabelling the covariant derivatives just in the same way as in the case of the linear
constraints (3.23).
It is worth mentioning a specific feature of the nonlinear chiral multiplet case as compared
with the linear multiplet case. The original world constraints (2.1)(a), (4.2) preserve the whole
automorphism SU(2)R group acting on the doublet indices of harmonics and Grassmann coordinates and break the PauliGrsey SU(2)PG symmetry realized on the doublet index a of q +a
down to a U (1) subgroup (due to the presence of the constant triplet cab in (4.2)). The same
symmetry structure is exhibited by the world constraints (4.5), (4.11) which are equivalent
to (4.2) and the analyticity condition (2.1)(a). Before fixing the frame gauge as in (4.13), the
superfields Qia are transformed by SU(2)R linearly, in the same way as the spinor derivatives
D i , D i , i.e., as
R Qia Qia (t, , ) Qia (t, , ) = ik Qka ,
(ik ) = ik ,
ii = 0.
(4.15)
Here ik are constant SU(2)R parameters. After imposing the gauge (4.13), this transformation
law becomes nonlinear, as it must be accompanied by the compensating gauge transformation
needed for preserving (4.13)
(1 i2 )comp = 12 + 11 ,
(1 + i2 )comp = 12 + 22 ,
(4.16)
whence
R = 11 + 212 + 22 ()2
(and c.c.),
(4.17)
i.e., and are transformed as projective CP 1 coordinates of the 2-sphere S 2 SU(2)R /U (1)R .
The constraints (4.14), where D i and D i are still transformed linearly with respect to their doublet indices and , are transformed according to the nonlinear transformation rule (4.17), are
directly checked to be SU(2)R covariant. This interpretation of the nonlinear chiral N = 4, d = 1
superfields as parameters of S 2 was the starting point of the derivation of the constraints (4.14)
187
in [13] (in an N = 4 superspace parametrization twisted as compared to ours). Note that both
and are inert under the U (1) remnant of the broken SU(2)PG symmetry.
Let us now discuss how the actions of the nonlinear chiral multiplet are reproduced from the
gauged q + actions.
The standard free action of q +a is obviously not invariant even under the rigid version of (4.1)
due to the presence of the rescaling isometry in (4.1). This situation is quite similar to what we
faced in [4] when deriving the action of the nonlinear (3, 4, 1) multiplet from the q + action with
a gauged rescaling invariance. It was shown there that the simplest invariant action is a nonlinear
action of the sigma-model type. In the case considered here we should start from the action which
is simultaneously invariant under the rescalings and the U (1) transformations. The unique object
invariant under both gauged isometries is constructed as
Y=
q +a cab q b
,
(q +a qa )
(4.18)
where
q a D q +a V1 q +a V2 cba q +b
(4.19)
V1,2
and
were defined in (3.12). It is the true nonlinear analog of the invariant X defined
in (3.35). It is easy to check that Y admits an equivalent representation in terms of the world
objects, such that it is manifestly invariant under the gauge transformations (4.7)
Q+a cab Qb
,
(Q+a Q
a)
and satisfies the relations
++ 2
2

D
Y = D Y = 0,

++
D Y D Y Y 2 = 1.
Y=
(4.20)
D ++ D Y = 2Y,
In the gauge (4.13) and frame (3.18)

+
2i
1
+
+
+
Y=
u2 u2 u1 u1 + (1 ) u1 u2 + u2 u1 .
2
(1 + )
A nonlinear analog of the relation (3.39) is

1 ++ 1 + +
+
= i Y u2 u2 D Y u2 u2 D Y u2 u2 .
2
2
1 +
(4.21)
(4.22)
(4.23)
From this relation and its conjugate it is easy to express and through Y and its harmonic
derivatives.
The general sigma-model type action of the nonlinear chiral multiplet corresponds to the
following gauged subclass of the general q +a actions

Schn = H L u, Y, D ++ Y, D Y .
(4.24)
After expressing Y in terms of the world objects, choosing the gauge (4.13), SU(2) frame
(3.18) and preforming the integration over harmonics, (4.24) becomes the general sigma model
action of the nonlinear chiral multiplet , as it was given in [18].
Let us point out that this correspondence, as in other similar cases [4,5], allows one to equivalently deal with the action in the original frame q + representation by choosing the appropriate
188
WessZumino gauge for the topological gauge superfields V1,2

and using the residual twoparameter gauge freedom to trade two out of four original physical bosonic fields of q +a for two
d = 1 gauge fields. The latter become just two auxiliary fields of the nonlinear chiral multiplet.
To work in the q + representation is in some aspects easier than to use the frame where the
same action looks as the general action of the superfields and .

It is interesting that in the case under consideration there also exists a unique SU(2)R invariant
action of the WZW type which is gauge invariant up to a total derivative in the integral and in
this sense is an analog of the free action of the linear chiral multiplet in the form (3.24), (3.25).
The quantity
F = q +a qa = e2v1 Q+a Q
a
(4.25)
is manifestly SU(2)R invariant and invariant under the gauge 2 transformations, while its 1
transformation reads
1 F = 21 F,
1 log F = 21 .
Then the action

Ssu(2) = H log F = H 2v1 + log Q+a Q
a
(4.26)
(4.27)
is gauge invariant since the full N = 4 superspace integral of the analytic parameter 1 vanishes.
Using (4.4), it is easy to check that the bridge v1 , modulo a constant and purely analytic term
which vanish after integration, in the gauge (4.13) is reduced to

+
+
+
v1 log 1 u+
(4.28)
1 u2 + 2u1 u1 + log 1 + u2 u1 + 2u2 u2 .
It follows from the constraints (4.14) that the integral of any holomorphic or antiholomorphic
function of , over the full N = 4 superspace vanishes. So the bridge v1 drops out from
(4.27) in the gauge (4.13) and, taking into account that in this gauge
Q+a Q
a = 1 + ,
we obtain the simple final expression for the action (4.27)

Ssu(2) = dt d 4 log(1 + ).
(4.29)
(4.30)
It describes the N = 4 superextension of the d = 1 sigma model on S 2 SU(2)R /U (1)R and

was derived in this form for the first time in [13]. The Lagrangian in (4.30) is just the corresponding Khler potential.3 Actually, using Eq. (4.23), we could write an equivalent representation for
Ssu(2) through the universal tensorial invariant Y . However, the SU(2)R symmetry in such a
representation is not manifest due to the presence of explicit harmonics over which one should
integrate.
Finally, we discuss the structure of two FI terms in the present case. They are originally given
++
actions (3.41) as in the linear case. Further, passing to the integral over the full
by the same V1,2
N = 4 superspace, using the relation

+a

+ icba Q+b u
v1 iv2 = log q +a + icba q +b u
(4.31)
a log Q
a
3 Formally, (4.30) looks also invariant under SU(2) , however the constraints (4.14) do not respect this second SU(2),
PG
so the latter is not a symmetry of (4.30).
189
and its conjugate (they follow from (4.4)), as well as the analyticity of q +a , and, finally, performing the integration over harmonics in the gauge (4.13), the FI terms can be expressed through ,
as follows

i
FI
S1 = 1 dt d 4 1 2 2 1 ,
2

1
FI
S2 = 2 dt d 4 1 2 + 2 1 .
(4.32)
2
like their customary chiral superfield analogs (3.43),
Surprisingly, they are still linear in , ,
(3.44). Note that the SU(2)R invariance of (4.32) can be checked with the help of the basic
constraints (4.14), taking into account that SU(2)R transforms the explicit s in (4.32) in the
standard way (rotates them in the doublet index), while and are transformed according to
the law (4.17) and its conjugate.
5. Non-Abelian gauge group
Let us now consider gauging of the last two-parameter group admitting a realization on the
analytic superfield q +a , the non-Abelian solvable group (2.13) consisting of a dilatation and a
shift of q +a . The gauge transformation laws are
q +a = 1 q +a + 2 u+a ,
(5.1)
where as before 1 and 2 are charge-zero analytic superfields. Due to the non-Abelian character of this group, its gauging is a little bit more tricky as compared to the previous two (Abelian)
cases. The commutation relations of the gauge transformations are given by
[, ]q +a = q +a ,
1 = 0,
2 = 2 1 1 2 .
(5.2)
In order to covariantize the harmonic constraints, we need to introduce two analytic gauge superfields V ++ , W ++ with the transformation laws
V ++ = D ++ 1 ,
W ++ = D ++ 2 + 1 W ++ 2 V ++ .
(5.3)
It is easy to check that the Lie bracket of two such transformations has the form (5.2). The
covariant harmonic constraint on the superfield q +a now reads
D ++ q +a V ++ q +a W ++ u+a = 0.
V ++
The gauge superfields

bridge superfields v, w as
V ++ = D ++ v,
and
W ++
(5.4)
are expressed through the corresponding non-analytic
W ++ = ev D ++ w.
(5.5)
As in other cases, the bridges v, w transform under two types of gauge transformations, the original ones with the analytic parameters 1 (, u) and 2 (, u), and new ones with the parameters
1 (t, , ) and 2 (t, , ) which are independent of harmonic variables:
v = 1 (, u) 1 (t, , ),
w = ev 2 (, u) 2 (t, , ) + 1 (t, , )w.
(5.6)
We also define the new non-analytic -world superfield

Q+a = ev q +a wu+a ,
Q+a = 1 Q+a + 2 u+a .
(5.7)
190
As a consequence of (5.4), the superfield Q+a is homogeneous in the harmonic variables

D ++ Q+a = 0
Q+a (t, , , u) = Qba (t, , )u+

b,
(Qab ) = Qab .
(5.8)
The harmonic-independent superfields Qba transform as

Qba = 1 Qba 2 ba .
(5.9)
The real parameter 2 may be used to gauge away the antisymmetric part of the tensor
Qab = Q(ab) .
Qba :
(5.10)
Then the target space scale invariance with the real parameter 1 may be used to fix the value of
a component of this tensor. We choose the gauge
Q21 = Q12 = i
+
Q+1 = Q11 u+
1 + iu2 ,
+
Q+2 = Q22 u+
2 + iu1 .
(5.11)
We now should take into account the analyticity constraints on the original superfields q +a .
When expressed in terms of the new superfields, these constraints become

D + Q+a + D + vQ+a + D + w + wD + v u+a = 0,
(5.12)
D + Q+a + D + vQ+a + D + w + w D + v u+a = 0.
Due to the analyticity of the gauge superfields V ++ , W ++ , the fermionic connections D + v,
D + w + wD + v have a simple dependence on harmonic variables
D ++ D + v = 0 D + v = Aa (t, , )u+
a,
+

++
+
+
D w + wD v = 0 D w + wD + v = B a (t, , )u+
D
a.
(5.13)
(5.14)
Analogously, for the conjugate connections we have

D + v = A a u+
a,
D + w + w D + v = B a u+
a.
(5.15)
Then, the constraints (5.12) imply, in ordinary superspace,

D 2 Q11 + A2 Q11 + iA1 B 1 = 0,
B 2 = iA2 ,
D 1 Q22 + A1 Q22 + iA2 + B 2 = 0,
B 1 = iA1 ,
D 1 Q11 + A1 Q11 = 0,
D 2 Q22 + A2 Q22 = 0.
(5.16)
(5.17)
From Eq. (5.16) one expresses the gauge connections

1
2iD 2 Q11 Q11 D 1 Q22 ,
11
22
4+Q Q

1
2iD 1 Q22 Q22 D 2 Q11 .
A2 = iB 2 =
11
22
4+Q Q
A1 = iB 1 =
(5.18)
Substituting these expressions into Eq. (5.17) and then using complex conjugation yield the full
set of the nonlinear constraints on the superfields Q11 , Q22 =

2iD 2 D 1 = 0,
4 +

2
2i D D 1 = 0,
D 1 +
4 +
D1 +
(5.19)

2 = 0,
2iD 1 D
4 +

2i D 1 D 2 = 0.
D 2 +
4 +
191
D 2 +
(5.20)
Eqs. (5.19) and (5.20) may be interpreted as (twisted) nonlinear chirality constraints on the su
perfields and its complex conjugate .
Let us now define the subclass of the general q + actions which respects the invariance under
the rigid transformations (2.13) and, after gauging, under their local counterparts (5.1). In the
frame it should yield the general sigma-model type action of the nonlinear chiral multiplet in
question.
The invariance under the shift transformations in (2.13) just means that the corresponding
ia
(ib) . Next, the
superfield Lagrangian cannot depend on the trace part in q ia u+
i , i.e., q q
invariance under the target space scale transformations constrains the action to depend only on
two independent ratios of three components of q (ab) , i.e., q (ab) q 11 /q 12 , q 22 /q 12 . In other
words, the appropriate general superfield q +a Lagrangian should be an arbitrary function of the
superfields q 11 /q 12 , q 22 /q 12 which can be interpreted as projective coordinates of some twosphere S 2 . Being reformulated in terms of the superfields q a , this requirement amounts to the
following particular choice of the q + Lagrangian

q a
L = L u ,
(5.21)
,
|q|
where

1
b +
q a = q a ua q +b u
b q ub ,
2
2 1 (ab)
+a + b 1 +a
|q|
2 = q +a q
= q q(ab) .
q ua + q a u+
a = q ua q ub
a
4
2
(5.22)
As in the previous case, the standard free action of q +a is not invariant under the target space
rescalings. The Lagrangians from the subclass (5.21) are of the sigma-model type, with nonconstant bosonic target metrics.
The gauging of these Lagrangians goes in the standard way, by subjecting q +a to the covariantized constraint (5.4) and defining q a in a gauge-covariant way as
q a = D q +a V q +a W u+a ,
q a = 1 q a + 2 ua .
(5.23)
The final gauge-covariantized action has the same form as the rigidly invariant one (5.21) but
with the superfields q a defined in a gauge-covariant way. Just due to this covariance, the basic
objects (5.22) admit the equivalent frame representation

1
b +
q a = ev Qa ua Q+b u
ev Q a ,
Q
u
b
b
2

b 1 +a

a + 2
Q
u
u
+
Q
u
Q
|q|
2 = e2v Q+a u+
a
a
a
b
4
1
2.
(5.24)
= e2v Q(ab) Q(ab) e2v |Q|
2
192
The covariantized superfield argument in (5.21) does not depend on the bridge v, whence

a
q a
Q
L = L u ,
(5.25)
.
= L u ,
|q|
|Q|
In the gauges (5.10) and (5.11):
Q 1 = Q1 = u
1 + iu2 ,

= 1 + ,
|Q|
2 = Q2 = u
+ iu ,
Q
2
1
(5.26)
and the action corresponding to the Lagrangian (5.25), after performing the integration over
Note that the relaharmonics, becomes the general off-shell action of the supermultiplet , .
tions (5.26), like analogous relations of the previous cases, are invertible:
1 +
= Q +1 u
2 Q u2 ,
2 u+ Q
+2 u .
= Q
1
1
(5.27)
=
This ensures the possibility to express , through the basic gauge invariant object, q a /|q|
and in fact proves the equivalence of the general N = 4 action of superfields ,
a /|Q|,
Q
and the particular class of the gauged q + actions defined above.
It is worth noting that the building blocks of the world gauge invariants can be successively
reproduced from the simplest invariant of the shift gauge transformation (with the parameter 2 )
+a +
ua ,
q ++ = q +a u+
a = q
q ++ = 1 q ++ .
(5.28)
Acting on (5.28) by the covariant derivative D V , we can produce new non-analytic

superfields which are invariant under the 2 transformations and covariant with respect to the
1 transformations:
q + =

1
D V q ++ , q = D V q + ,
2
q = 1 q .
(5.29)
They are related to the superfields q a and q a by

1 +a
+a
a
ua = q a u+
q = q a u
ua
q ua + q a u+
a = q
a,
a = q
2
and can be used to form two independent gauge invariant ratios
q + =
X ++ =
q ++
q ++ q (q + )2
X =
q
q ++ q (q + )2
(5.30)
(5.31)
which are just independent harmonic projections of the superfield argument in (5.25).
Let us now dwell on the peculiarities of the realization of SU(2)R and SU(2)PG symmetries
on the superfields and and the surprising relation to the nonlinear chiral multiplet discussed
in the previous Section.
The basic gauge covariant constraint (5.4) clearly breaks the original SU(2)R SU(2)PG symmetry realized on q +a , Grassmann and harmonic coordinates down to the diagonal R-symmetry
group SU(2)R which uniformly rotates all doublet indices. The gauge (5.10) is SU(2)R covariant,
so the superfield Q(ab) is transformed as
R Q(ab) Q(ab) (t, , u ) Q(ab) (t, , u) = ad Q(db) + bd Q(ad) ,
bb = 0.
(5.32)
193
The gauge (5.11) is not preserved under (5.32), and in order to restore this gauge one should
accompany the SU(2)R transformations by a compensating 1 transformation with

(1 )comp = i 22 11 .
(5.33)
As a result, in this gauge the superfields and are nonlinearly transformed under SU(2)R
+ i22 ()2 ,
R = 212 i11 (2 + )
R = (R ),
(5.34)
and so can be treated as coordinates of the coset S 2 SU(2)R /U (1)R in a particular parametrization. Obviously, there should exist an equivalence transformation to the stereographic
projection parametrization in which the S 2 coordinates are transformed according to the holomorphic law (4.17). The precise form of this field redefinition is as follows
=i

,
1 + 1 +
= 2i
R = 11 + 212 + 22 ()2
,
1
(5.35)
(and c.c.).
(5.36)
The transformation law (5.36) coincides with (4.17), which suggests that in this new holomorphic
parametrization the constraints (5.19), (5.20) take the form (4.14). Indeed, a simple calculation
shows that after the field redefinition (5.35) the constraints (5.19), (5.20) are equivalently rewritten as
D 1 + D 2 = 0,
D 1 + D 2 = 0 (and c.c.).
(5.37)
Thus we see that the nonlinear chiral multiplet considered in this Section is in fact a disguised
form of the nonlinear (twisted) chiral multiplet of Ref. [13] rederived within the gauging procedure in Section 4. This is rather surprising, because in the two cases we gauged two essentially
different two-parameter groups, respectively, Abelian and non-Abelian ones (2.12)(b) and (2.13).
The identity of these two multiplets amounts to the identity of their general actions, despite the
fact that the classes of the appropriate q + actions one starts with in these two cases are essentially
different. Here we again encounter the phenomenon of non-uniqueness of the inverse oxidation
procedure as compared with the target space dimensional reduction [4,5]: the same off-shell multiplet can be recovered by gauging some non-equivalent isometries of the root multiplet. For
instance, the N = 4, d = 1 multiplet (1, 4, 3) and its most general action can be obtained from
the q + multiplet and the appropriate set of the q + actions by gauging either the non-Abelian
SU(2)PG group or the Abelian group of three independent shift isometries of q +a [5]. Basically,
the difference between these two gauging procedures lies only in the fact that they start from
different subclasses of the general set of q + actions. However, the final action of the reduced
multiplet does not remember from which parent q + action it originated.
Taking for granted that all off-shell N = 4, d = 1 superfields can be recovered from the q +a
superfield by gauging different symmetries realized on the latter and taking into account that
only three independent two-parameter groups (defined in (2.12) and (2.13)) can be implemented
on q +a , we conclude that only two essentially different off-shell N = 4, d = 1 multiplets with
the content (2, 4, 2) exist: the standard linear chiral multiplet and the nonlinear chiral multiplet
introduced in [13]. Any other version of the chiral multiplet should be reducible to one of these
two via some field redefinition.
There is one more way to see that the constraints (5.19), (5.20) are equivalent to (4.14). After some algebra, using (5.19), (5.20) at the intermediate steps, the expressions for the spinor
194
connections (5.18) can be cast in the following form

i
1
2

D 1 log 1 + 1 + ,
A =D
1 + 1 +

i
2
1

D 2 log 1 + 1 + .
A =D
1 + 1 +
(5.38)
Substituting these expressions and their complex conjugates into (5.17) and complex conjugates
of (5.17), we recover (5.37), with being related to just by Eq. (5.35).
Finally, as an instructive example, we present the SU(2)R invariant action in terms of the
original superfield variables, as well as the relevant FI term.
The SU(2)R invariant action is given by an expression similar to (4.27)

.
Ssu(2) = H log |q|
(5.39)
= H v + log |Q|
It is manifestly invariant under the gauge shift 2 transformation (since q +a is invariant), as well
as under the scale and shift gauge transformations. It is also invariant under the scale gauge
1 transformations since under the latter the Lagrangian in (5.39) is shifted by an analytic gauge
parameter the integral of which over the full N = 4 superspace vanishes:

1 log |q|
(5.40)
= 1 ,
H 1 = 0.
we
To find the precise form of the action in terms of the nonlinear chiral superfields , ,
should make use of Eq. (5.26) and also compute the bridge part of the N = 4 superspace integral
in (5.39):

H v.
(5.41)
This integral can be evaluated by taking one spinor derivative, say D + , off the measure d 4 ,
throwing it on v, expressing D + v as in (5.13), doing the harmonic integral du, substituting the
gauge-fixed expressions (5.38) for the spinor connections A1 , A2 and, finally, restoring the full
Grassmann measure by taking the spinor derivatives D 1 , D 2 off these expressions. It turns out
that only the second terms in the expressions (5.38) contribute, and we obtain

H v = dt d 4 log 1 + 1 + .
(5.42)
Using this in (5.39), we obtain

Ssu(2) = dt d 4 log 1 + log 1 + 1 + .
(5.43)
Now it is straightforward to check that, after passing to the superfields , via (5.35), the Lagrangian in (5.43) is reduced (modulo a constant shift) just to
log(1 + ).
Thus we obtain the expected result that the action (5.39), (5.43) is in fact identical to the previously considered SU(2)R invariant action (4.30).
195
As for the FI terms, in the present case only the gauge superfield V ++ possesses an Abelian
gauge transformation law, so one is able to construct only one FI term:

FI
(2) ++
V
= iv H + + + v.
Sv = iv du d
(5.44)
Inserting the identities 1 = D + , 1 = D + into the round brackets in the r.h.s. of (5.44),
integrating by parts with respect to spinor derivatives, using the relations (5.13) with (5.38) and
their conjugates, doing harmonic integral and, at the end, integrating by parts once again, one
finally finds

SvFI = iv dt d 4 1 2 2 1 ,
(5.45)
which coincides with one of the FI terms in (4.32).
6. Conclusions
In this article and two previous papers [4,5] we showed that all known off-shell N = 4,
d = 1 multiplets with 4 physical fermions can be reproduced from the basic (root) multiplet (4, 4, 0) by gauging some symmetries, Abelian or non-Abelian, realized on this multiplet.
The corresponding general N = 4 mechanics actions are recovered as the result of the proper
gauge-fixing in the appropriate gauged subclasses of the general q + action, the subclasses which
enjoy invariance under the symmetries just mentioned. Our gauging procedure uses the manifestly supersymmetric universal language of N = 4 superspace and does not require to resort to
component considerations at all. Another merit of our approach is that it reduces the whole set of
non-equivalent superfield actions of the N = 4 mechanics models to some particular cases of the
generic q + action extended by non-propagating topological gauge superfields. Just the presence of the latter enables one to preserve the manifest supersymmetry at each step and to reveal
the irreducible off-shell superfield contents of one or another model by choosing the appropriate
superfield gauges and (in the cases considered in the present paper) by passing to the equivalent
frame formulations. The alternative (and in many cases more technically feasible) way of doing suggested by the gauging approach is to always stay in the initial q + representation where
the harmonic analyticity is manifest and to choose the WZ gauge for the relevant analytic nonpropagating gauge superfields. Each topological gauge multiplet in the WZ gauge contributes
just one scalar (gauge) field which, after fully fixing the residual gauge freedom, becomes
an auxiliary field of the new off-shell N = 4 multiplet related to the q + multiplet via linear
or nonlinear versions of the automorphic duality [6]. Thus in the component formulation our
approach automatically yields the explicit realization of this intrinsically one-dimensional offshell duality. The distinctions between various types of this duality are related to the differences
between the global symmetry groups subjected to gauging.
The basic peculiarity of the cases considered in this paper as compared to those treated in
[4,5] is that the superfields describing the N = 4 multiplets (2, 4, 2) do not live on the N = 4
analytic harmonic subspace (as distinct from the multiplets (0, 4, 4), (1, 4, 3) and (3, 4, 2)). They
are most naturally described after passing to the equivalent frame [16,17], with the ordinary
N = 4 superfield gauge parameters and the harmonic superfield bridges to the frame as the
basic gauge objects. These bridges ensure the equivalence of the manifestly analytic frame picture one starts with and the picture in the frame. In the frame, the original gauge-covariantized
196
analyticity-preserving harmonic constraints on the superfield q +a amount to the harmonic independence of the involved superfields. The harmonic Grassmann analyticity, which is manifest in
the frame, in the frame amounts to the covariant analyticity conditions. After properly fixing
N = 4 supersymmetric gauges, these conditions become the linear or nonlinear N = 4 chirality conditions, depending on which two-parameter symmetry group realized on q +a is subjected
to gauging. There are only three such groups and they are listed in (2.12) and (2.13). We considered gauging of all these three groups and found that the gauging of the group (2.12)(a) leads
to the linear chiral N = 4 multiplet, while gaugings of (2.12)(b) and (2.13) lead to the same
nonlinear chiral multiplet [13], despite the obvious non-equivalence of these two groups. This
non-uniqueness is a manifestation of the general non-uniqueness of the oxidation procedure as
inverse to the automorphic duality. Since only three two-parameter symmetries can be realized
on q +a , from our results it follows, in particular, that no other non-equivalent nonlinear chiral
N = 4 multiplet can be defined.
Interesting venues for further applications of our gauge approach are provided by models of
N = 8 supersymmetric mechanics (see, e.g., [19,20] and refs. therein). It was argued in [20],
by considering a wide set of examples, that the off-shell N = 8 multiplet (8, 8, 0) is the true
N = 8 analog of the root N = 4 multiplet (4, 4, 0) and that the whole set of the component
actions of the N = 8 mechanics models with 8 physical fermions (and finite numbers of auxiliary fields) follow from the general action of this basic N = 8 multiplet via a linear version of
the automorphic duality. It would be interesting to apply our techniques to these cases. Recall
that our approach is bound by the requirement that the symmetries to be gauged commute with
supersymmetry. In the N = 8 case the target space scale and shift transformations still obey this
criterion, so one can hope that the gauging would nicely work in this case too and could help
to understand the relationships between the multiplet (8, 8, 0) and the rest of the N = 8 multiplets in a manifestly N = 4 supersymmetric superfield fashion. We can also hope to discover in
this way new nonlinear N = 8 multiplets and the corresponding new N = 8 mechanics models,
besides those already known [14,15]. The primary question to be answered is how to define an
N = 8 analog of the N = 4 topological gauge multiplet which plays a crucial role in our approach. Another possible way of extending our study is to construct topological N = 4, d = 1
supergravity multiplets and to gauge, with their help, the R-symmetry SU(2) groups of N = 4
supersymmetry, with new models of N = 4 mechanics as an outcome. Finally, let us note that
the nonlinear chiral multiplets exist also in dimensions d>1 [13], e.g., in d = 3 [21]. It would be
interesting to inquire whether they can also be derived by gauging some symmetries realized on
the appropriate analytic harmonic superfields, some analogs of q +a , i.e., whether their defining
constraints are also a disguised frame form of the harmonic analyticity conditions.
Acknowledgements
The work of E.I. was supported in part by the RFBR grant 06-02-16684, the RFBR-DFG grant
06-02-04012-a, the grant DFG, project 436 RUS 113/669/0-3, the grant INTAS 05-7928 and a
grant of Heisenberg-Landau program. He thanks Laboratoire de Physique, UMR5672 of CNRS
and ENS Lyon, for the warm hospitality extended to him during the course of this work.
References
[1] R. de Lima Rodrigues, The quantum mechanics SUSY algebra: An introductory review, hep-th/0205017;
A.V. Smilga, Low dimensional sisters of SeibergWitten effective theory, hep-th/0403294.
197
[2] S. Bellucci, S.J. Gates Jr., E. Orazi, A Journey Through Garden Algebras, Lectures given at Winter School on
Modern Trends in Supersymmetric Mechanics (SSM05), Frascati, Italy, 712 March 2005, hep-th/0602259.
[3] Z. Kuznetsova, M. Rojas, F. Toppan, JHEP 0603 (2006) 098, hep-th/0511274;
F. Toppan, Irreps and Off-shell Invariant Actions of the N-extended Supersymmetric Quantum Mechanics, Fifth
International Conference on Mathematical Methods in Physics IC2006, CBPF, Rio de Janeiro, Brazil, 2428 April
2006, hep-th/0610180.
[4] F. Delduc, E. Ivanov, Nucl. Phys. B 753 (2006) 211, hep-th/0605211.
[5] F. Delduc, E. Ivanov, Nucl. Phys. B 770 (2007) 179, hep-th/0611247.
[6] S.J. Gates, Jr., L. Rana, On Extended Supersymmetric Quantum Mechanics, Maryland Univ. Preprint # UMDPP
93-24, October 1994;
S.J. Gates Jr., L. Rana, Phys. Lett. B 342 (1995) 132, hep-th/9410150;
S.J. Gates Jr., L. Rana, Phys. Lett. B 345 (1995) 233, hep-th/9411091;
A. Pashnev, F. Toppan, J. Math. Phys. 42 (2001) 5257, hep-th/0010135.
[7] S. Bellucci, S. Krivonos, A. Marrani, E. Orazi, Phys. Rev. D 73 (2006) 025011, hep-th/0511249.
[8] S. Hellerman, J. Polchinski, Supersymmetric quantum mechanics from light cone quantization, hep-th/9908202.
[9] F. Delduc, S. Krivonos, 2005, unpublished.
[10] C. Burdik, S. Krivonos, A. Shcherbakov, Czech. J. Phys. 55 (2005) 1357, hep-th/0508165.
[11] S. Krivonos, A. Shcherbakov, Phys. Lett. B 637 (2006) 119, hep-th/0602113.
[12] E. Ivanov, O. Lechtenfeld, JHEP 0309 (2003) 073, hep-th/0307111.
[13] E. Ivanov, S. Krivonos, O. Lechtenfeld, Class. Quantum Grav. 21 (2004) 1031, hep-th/0310299.
[14] E. Ivanov, Phys. Lett. B 639 (2006) 579, hep-th/0605194;
S. Bellucci, S. Krivonos, A. Marrani, Phys. Rev. D 74 (2006) 045005, hep-th/0605165.
[15] S. Bellucci, A. Beylin, S. Krivonos, A. Shcherbakov, Phys. Lett. B 633 (2006) 382, hep-th/0511054.
[16] A. Galperin, E. Ivanov, V. Ogievetsky, E. Sokatchev, Pisma Zh. Eksp. Teor. Fiz. 40 (1984) 155, JETP Lett. 40
(1984) 912;
A.S. Galperin, E.A. Ivanov, S. Kalitzin, V.I. Ogievetsky, E.S. Sokatchev, Class. Quantum Grav. 1 (1984) 469.
[17] A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E.S. Sokatchev, Harmonic Superspace, Cambridge Univ. Press, 2001,
306 p.
[18] S. Bellucci, A. Beylin, S. Krivonos, A. Nersessian, E. Orazi, Phys. Lett. B 616 (2005) 228, hep-th/0503244.
[19] S. Bellucci, E. Ivanov, S. Krivonos, O. Lechtenfeld, Nucl. Phys. B 699 (2004) 226, hep-th/0406015.
[20] E. Ivanov, O. Lechtenfeld, A. Sutulin, Hierarchy of N = 8 Mechanics Models, Hannover University Preprint ITPUH-11/07, arXiv: 0705.3064 [hep-th].
[21] S. Bellucci, S. Krivonos, A. Shcherbakov, Phys. Rev. D 74 (2006) 065016, hep-th/0606052.
Testing neutrino masses in little Higgs models

via discovery of doubly charged Higgs at LHC
A. Hektor, M. Kadastik, M. Mntel, M. Raidal, L. Rebane
National Institute of Chemical Physics and Biophysics, Ravala 10, Tallinn 10143, Estonia
Abstract
We have investigated the possibility of direct tests of little Higgs models incorporating triplet Higgs
neutrino mass mechanism at LHC experiments. We have performed Monte Carlo studies of DrellYan pair
production of doubly charged Higgs boson ++ followed by its leptonic decays whose branching ratios
are fixed from the neutrino oscillation data. We propose appropriate selection rules for the four-lepton
signal, including reconstructed taus, which are optimized for the discovery of ++ with the lowest LHC
luminosity. As the Standard Model background can be effectively eliminated, an important aspect of our
study is the correct statistical treatment of the LHC discovery potential. Adding detection efficiencies and
measurement errors to the Monte Carlo analyses, ++ can be discovered up to the mass 250 GeV in the
first year of LHC, and 700 GeV mass is reachable for the integrated luminosity L = 30 fb1 .
1. Introduction
The main motivation of the Large Hadron Collider (LHC) experiment is to reveal the secrets
of electroweak symmetry breaking. If the light Standard Model (SM) Higgs boson H will be
discovered, the question arises what stabilizes its mass against the Planck scale quadratically divergent radiative corrections. The canonical answer to this question is supersymmetry, predicting
a very rich phenomenology of sparticles in the future collider experiments.
Alternatively, the light SM Higgs boson may signal some strong dynamics at high scale
4f , where f is the decay constant of the new strongly interacting theory [1]. The most
E-mail address: liis.rebane@cern.ch (L. Rebane).

doi:10.1016/j.nuclphysb.2007.07.014
A. Hektor et al. / Nuclear Physics B 787 (2007) 198210
199
interesting class of models in such a scheme are the little Higgs models [24]. In those models the SM Higgs boson is a pseudo Goldstone mode of a broken global symmetry and remains
much lighter than the other modes of the model, thus solving the little hierarchy problem and
postponing the solution to the fundamental hierarchy problem to the scale . Those models are
also very interesting from collider physics point of view since they predict the existence of new
particles, such as a new set of heavy gauge bosons WH , ZH , a vectorlike heavy quark pair T , T
with charge 2/3, and triplet Higgs bosons . If the new particle masses are O(1) TeV, direct
tests of the models are possible at LHC [57].
An important open issue to address in the context of little Higgs models is the origin of nonzero neutrino masses [812]. The neutrino mass mechanism which naturally occurs in those
models is the triplet Higgs mechanism [13,14] (sometimes called type II seesaw) which employs
a scalar with the SU(2)L U (1)Y quantum numbers (3, 2). The existence of such a multiplet
in some versions of the little Higgs models is a direct consequence of global symmetry breaking
which makes the SM Higgs light. For example, in the minimal littlest Higgs model [15], the
triplet Higgs with non-zero hypercharge arises from the breaking of global SU(5) down to SO(5)
symmetry as one of the Goldstone bosons. Its mass M gs f , where gs < 4 is a model
dependent coupling constant in the weak coupling regime [1], is therefore predicted to be below
the cut-off scale , and could be within the mass reach of LHC. Although the triplet mass scale
is O(1) TeV, the observed neutrino masses can be obtained naturally. Firstly, non-observation of
rare decays eee, e , , where = e, , implies that the triplet Higgs boson
Yukawa couplings Yij must be small, thus suppressing also the neutrino masses. Secondly, the
vaccuum expectation value (vev) of the neutral component of triplet v contributes at tree level
to the SM oblique corrections, and is therefore severely constrained by precision data. There exist
additional mechanisms which can explain the smallness of v in little Higgs models. Since the
smallness of v is the most natural explanation of the smallness of neutrino masses in the little
Higgs models, we assume this to be the case in this work.
The aim of this paper is to study the possibility of direct tests of little Higgs models and
neutrino mass mechanisms at LHC experiments via pair productions and subsequent decays of
triplet Higgs boson. We study the DrellYan pair production of doubly charged component of
the triplet [1621]
pp ++ ,
(1)
followed by the leptonic decays. Notice that (i) the production cross section does not depend
on any unknown model parameter but the mass of ++ ; (ii) smallness of v in this scenario,
following from the smallness of neutrino masses, implies that the decays ++ W + W + are
negligible, and we neglect this channel in the following analyses; (iii) the ++ leptonic decay
branching fractions do not depend on the size of the Yukawa couplings but only on their ratios
which are known from neutrino oscillation experiments. In the triplet model the normally hierarchical light neutrino masses predict BR( ++ + + ) BR( ++ + + ) BR( ++
+ + ) 1/3. Therefore this scenario is predictive and testable at LHC experiments.
The production process (1) has been studied before in various theory papers. In this work we
first carry out a pure Monte Carlo study of the signal and background processes in the environment of LHC detectors. After that we improve our analyses by adding particle reconstruction
efficiencies and Gaussian distortion functions for particle momentas and ETmiss . Those mimic the
detector inefficiency effects at the Monte Carlo level. We believe that those effects help us to
estimate the realistic mass reach of the LHC detectors to the process under study.
200
In our study the new results are the following. For the signal reconstruction we use new
criteria, such as equality of invariant masses of positively and negatively charged leptons together
with total pT cut for all leptons, which allows us to achieve better reconstructions efficiencies
compared to the standard cuts. We also reconstruct tau lepton final states with more than one ,
which has not done before in this context. As all the SM background can be eliminated in the
case of this process, correct statistical analyses of the results in the limit of no background is an
important aspect of our study. For the discovery criteria we have used the Log-Likelihood Ratio
(LLR) statistical method to demand 5 discovery potential to be bigger than 95% (1 CLs+b >
0.95). Our results are optimized for the discovery of process (1) with the lowest possible LHC
luminosity. The pure Monte Carlo study shows that ++ up to the mass 300 GeV is reachable
in the first year of LHC (L = 1 fb1 ) and ++ up to the mass 800 GeV is reachable for the
luminosity L = 30 fb1 . Including the Gaussian measurement errors in the Monte Carlo the
corresponding mass reaches become 250 GeV and 700 GeV, respectively. The errors of our
estimates of the required luminosity for discovery depend strongly on the size of statistical Monte
Carlo sample of the background processes.
The paper is organized as follows. In Section 2 we present the collider phenomenology of
triplet Higgs boson and relate collider observables to neutrino mass measurements. In Section 3
we discuss the Monte Carlo produced signal and background processes. In Section 4 we present
the details of reconstruction and analysis procedure and results. Detector effects are discussed in
Section 5. Finally we conclude in Section 6.
2. Neutrino masses and collider phenomenology
In this work we consider little Higgs scenarios in which, due to the breaking of global symmetry protecting the SM Higgs boson mass, the spectrum of the model contains also a pseudo
Goldstone boson with the SU(2)L U (1)Y quantum numbers (3, 2) [15,22]. Although
is predicted to be heavier than the SM Higgs boson, the little Higgs philosophy implies that its
mass could be O(1) TeV [1]. Due to the specific quantum numbers the triplet Higgs boson couples only to the left-chiral lepton doublets Li (2, 1), i = e, , , via the Yukawa interactions
given by
L = i L ci 2 Y ij ( )Lj + h.c.,
(2)
where Yij are the Majorana Yukawa couplings. The interactions (2) induce lepton flavour violating decays of charged leptons which have not been observed. The most stringent constraint
on the Yukawa couplings comes from the upper limit on the tree-level decay eee and
is1 Yee Ye < 3 105 (M/TeV)2 [18,23]. Experimental bounds on the tau Yukawa couplings
are much less stringent. In our collider studies we take Y = 0.01 and rescale other Yukawa
couplings accordingly. In particular, hierarchical light neutrino masses imply Yee , Ye Y
consistently with the direct experimental bounds.
According to Eq. (2), the neutral component of the triplet Higgs boson 0 couples to the lefthanded neutrinos with the same strength as ++ couples to the charged leptons. If 0 acquires a
vev v , non-zero Majorana masses are generated for the left-handed neutrinos [13,14]. Non-zero
neutrino masses and mixing is presently the only experimentally verified signal of new physics
1 In little Higgs models with T -parity there exist additional sources of flavour violation from the mirror fermion sector
[24,25].
201
beyond the SM. In the triplet neutrino mass mechanism the neutrino masses are given by
(m )ij = Yij v .
(3)
We assume that the smallness of neutrino masses is explained by the smallness of v . In a realistic scenario massless Majoron, the Goldstone boson of broken lepton number, must be avoided.
This is achieved by an explicit coupling of to the SM Higgs doublet H via 0 H 0 H 0 [14],
where has a dimension of mass. If M , in the concept of seesaw [26] the smallness of
2,
neutrino masses is attributed to the very high scale of triplet mass M because v = v 2 /M
where v = 174 GeV. However, in the little Higgs models the triplet mass scale O(1) TeV alone
cannot suppress v . Therefore in this model M , which can be achieved, for example, via
shining of explicit lepton number violation from extra dimensions as shown in Refs. [27,28], or
if the triplet is related to the Dark Energy of the Universe [29,30]. Models with additional (approximate) T -parity [22] make the smallness of v technically natural. However, if the T -parity
is exact, v must vanish. In this work we do not consider the naturalness criteria and assume that
the above described neutrino mass scenario is realized in nature. In that case Y v O(0.1) eV
while the Yukawa couplings Y can be on the order of charged lepton Yukawa couplings of the
SM. As a result, the branching ratio of the decay W W is negligible. We also remind that
v contributes to the SM oblique corrections, and the precision data fit T < 2 104 [31] sets
an upper bound v 1.2 GeV on that parameter.
Notice the particularly simple connection between the flavour structure of light neutrinos and
the Yukawa couplings of the triplet via Eq. (3). Therefore, independently of the overall size of the
Yukawa couplings, one can predict the leptonic branching ratios of the triplet from neutrino oscillations. For the normally hierarchical light neutrino masses neutrino data implies negligible
branching fractions to electrons and BR( ++ + + ) BR( ++ + + ) BR( ++
+ + ) 1/3. Those are the final state signatures predicted by the triplet neutrino mass mechanism for collider experiments.
At LHC ++ can be produced singly and in pairs. The cross section of the single ++
2 . In the context of
production via the W W fusion process [18] qq q q ++ scales as v
++
+
W W + , was studied in
the littlest Higgs model this process, followed by the decays
Refs. [5,7,32]. The detailed ATLAS simulation of this channel shows [32] that in order to observe
an 1 TeV ++ , one must have v > 29 GeV. This is in conflict with the precision physics
bound v 1.2 GeV as well as with the neutrino data. Therefore the W W fusion channel is not
experimentally promising for the discovery of doubly charged Higgs.
On the other hand, the DrellYan pair production process pp ++ is not suppressed
by any small coupling and its cross section is known up to next to leading order [19] (possible additional contributions from new physics such as ZH are strongly suppressed and we neglect those
effects here). Followed by the lepton number violating decays , this process allows
to reconstruct invariant mass from the same charged leptons rendering the SM background
to be very small in the signal region. If one also assumes, as we do in this work, that neutrino
masses come from the triplet Higgs interactions, one fixes the leptonic branching ratios.
This allows to test the triplet neutrino mass model at LHC.
3. Monte Carlo simulation of the signal and backgrounds
The production of the doubly-charged Higgs is implemented in the PYTHIA Monte Carlo
generator [33]. The final and initial state interactions and hadronization have been taken into
account. We have used the CTEQ5L parton distribution functions.
202
Table 1
Cross-sections, numbers of Monte Carlo generated events and the corresponding integrated luminosities of the generated events. For the signal events we have taken the branching ratios BR( ) = BR( ) =
BR( ) = 1/3
Process
Total
(fb)
N of events
generated
Corresponding
luminosity (fb1 )
Signal
M = 200 GeV
M = 500 GeV
M = 1000 GeV
7.78E+01
1.99E+00
5.58E02
1.00E+05
1.00E+05
1.00E+05
1.28E+03
5.03E+04
1.79E+06
Background
pp t t 4
pp t t Z
pp ZZ
8.84E+04
6.50E+02
2.12E+02
2.55E+07
1.50E+05
1.00E+05
2.88E+02
2.3E+02
4.72E+02
In the following analysis the normal hierarchy of neutrino masses and a very small value of
the lowest neutrino mass is assumed. Such a model predicts that Higgs decay into electrons can
be neglected and that there are three dominant decay channels for ++ with approximately equal
branching ratios:
,
,
.
We have studied only pair production of doubly charged Higgs due to the reasons pointed out
above. pair decay products can combine to five different and combinations: 4, 31 ,
22 , 13 and 4 . Before reaching the detector, decays into an e, or a hadronic jet (marked
as j below) with branching ratios of 0.18, 0.17 and 0.65, respectively [34]. hadronic jets and s
are well visible and reconstructible in detector. The reconstruction of an energetic from electron
decay is sensitive to detector effects, involving sophisticated background processes [35]. In the
current analyses we will neglect this channel, which will cause 31% loss of the total signal. Such
loss is still sufficiently low and can be considered acceptable. Table 1 gives the cross sections
and the Monte Carlo generated event numbers in our study.
The signatures of decay are very clean due to (i) high transfer momentum of the decay
products, (ii) lepton number violation and (iii) pair production of . The Standard Model particles are lighter than , so the background s and s must have smaller transverse energy and
they do not produce an invariant mass peak in + + , + + , + + , final states. The present
lower bound for the invariant mass of is set by Tevatron to M 136 GeV [36,37]. In our
study, four-lepton background processes with reasonable cross sections and high pT leptons arise
from three Standard Model processes
pp t t,
pp t tZ,
pp ZZ.
PYTHIA was used to generate t t and ZZ background (t t is forced to decay to W W bb and
W leptonically). The CTEQ5L parton distribution functions were used. CompHEP was used to
203
generate the Zt t background via its PYTHIA interface [38,39]. All the datasets were generated
in Baltic Grid. In addition to background processes shown in Table 1, some other four-lepton
As
background processes exist involving b-quarks in the final state (for example, pp bb).
such processes are very soft, it is possible to use the effective tagging methods [40] and totally
eliminate this soft background [41]. Also, we do not consider possible background processes
from the physics beyond the Standard Model.
4. Reconstruction and analysis of the Monte Carlo data
To study the feasibility of detecting the signal over background, we have to work with five
possible reconstruction channels according to the following final states.
++ 4: The cleanest and most simple channel.
++ 31 : The channel is easily reconstructable using an assumption that the neutrino originating from the decay is collinear with -jet and gives majority to the missing
transverse energy (ETmiss ).
++ 22 : The signature can be reconstructed using the same assumptions for both
-neutrinos. The whole ETmiss vector has to be used here, while in the previous channel only
one component was needed.
++ 13 : The channel can be reconstructed theoretically relying on an additional
requirement that the two Higgs bosons have equal invariant masses. However, the reconstruction is very sensitive to the experimental accuracy of ETmiss determination.
++ 4 : The channel cannot be reconstructed (and triggered by the single muon
trigger).
First, we apply general detector related cut-offs for the Monte Carlo generated data. Generated particles were reconstructed within the pseudorapidity region || < 2.4 and with transverse
momentum higher than 5 GeV. These are the natural restrictions of the CMS and ATLAS detectors at the LHC. Only the pseudorapidity region || < 2.4 is reachable for the detector and only
the events with pT > 5 GeV are typically triggered. These restrictions suppress mainly the soft
Standard Model background. The efficiency of lepton reconstruction and charge identification
rate are very high, we use the values 0.9 and 0.95, respectively [42].
The invariant mass of two like-sign s and/or s are calculated using equation:
m
I
2
2

= p1 + p2 ,
(4)
where p1,2 is the or 4-momentum. Since the like-sign signal of s or s originate from
a doubly charged Higgs boson, the invariant mass peak measures the mass of doubly charged
Higgs, mI = M . 4- final state allows to obtain invariant masses directly from Eq. (4). In
channels involving one or several s, which are registered as -jets or secondary s (marked as
below), the momenta of jets has to be corrected according to the equation system:
pi = k i pijet ,

pT miss =
piT ,
(5)
(6)
M ++ = M ,
(7)
204
Table 2
Probabilities of all possible decay chains for pairs in our scenario. x in the table marks or, after decay, -jet.
The reconstructed signatures are marked in bold, the remaining signatures were not reconstructed. After omitting the
channels that include e decay, 69% of the total signal is left. In total 64% of the signal has been reconstructed
Decay channel
After decay (x = jet )
After Higgs
decay (x = )
0 ( )
1 ( )
2 ( )
3 ( )
4 ( )
2 4
2 31x
2 22x
2 13x
2 4x
0.1111
0.2222
0.3333
0.2222
0.1111
0.1111
0.1443
0.1407
0.0610
0.0198
0.0377
0.0736
0.0478
0.0207
0.0107
0.0125
0.0054
0.0012
0.0014
0.0001
Sum
1.0
0.64 + 0.05
where i counts s, p marks 3-momentum, pT is the vector of transverse momentum of the

produced neutrinos, pT miss is the vector of missing transverse momentum (measured by the detector) and ki > 1 are positive constants. Eq. (5) describes the standard approximation that the
decay products of a heavily boosted are collinear [42]. Eq. (6) assumes missing transverse
energy only to be comprised of neutrinos from decays. In general, it is not a high-handed
simplification, because the other neutrinos in the event are much less energetic and the detector
error of ETmiss is order of magnitude smaller [43]. Using the first two formulas, it is possible to
reconstruct up to two s per event. Additional requirement of Eq. (7) allows to reconstruct the
third per pair event, although very low measurement errors are needed.
A significant fraction of s (0.18) decay into s that cannot be distinguished from primary
s in the detector. Still, if reconstructed invariant masses of ++ and are considerably
different, we can suspect that one or several s originate form decays. In such case we can
again use Eqs. (5)(7) to correct the 4-momenta of decay products. When only one secondary
muon is present, ETmiss points into the same direction as its pT . Otherwise ETmiss is a superposition
of neutrino transverse momenta. Such correction tightens the invariant mass peak of the signal
and does not produce any artificial background.
The occurrence probability of different reconstruction channels are presented in Table 2. The
second column shows probabilities of Higgs decay to N s and M s. Next columns describe
the final state after decay to s and/or jets. Different columns mark the number of secondary
s and the rows designate -jets in the detector recordings. The events having at least one
e in a final state are omitted in our analysis, as well as events with M > 3 or N ( ) > 2.
The proportions of reconstructible signatures are marked in a bold-face. The table shows that 03
jet channels together with correction are almost equally important and overall reconstructible
channels comprise 64% of total events.
A clear signal extraction from the Standard Model background can be achieved using a set of
selection rules imposed on a reconstructed event in the following order.
S1: events with at least 2 positive and 2 negative muons or jets which have || < 2.4 and
pT >5 GeV are selected.
S2:
pT (scalar) sum of 2 most energetic positive and negative s or -jets has to be
bigger than a certain value (depending on Higgs mass).
S3: Z-taggingif invariant mass of the pair of opposite charged s or -jets is nearly equal
to Z mass (8595 GeV), then the particles are eliminated from the analysis.
205
S4: as s are produced in pairs, the reconstructed invariant masses (in one event) have to
be equal. We have used the condition
0.8 < m++

I /mI < 1.2.
(8)
If the invariant masses satisfy the condition then we include them to the histogram, otherwise we suspect that some s may originate from decay, and make an attempt to find
corrections to their momenta according to the method described above.
The rule S1 is an elementary detector trigger. S2, performing scalar sum of pT , is an untraditional cut. The advantage compared to the widely used pT cut for a single particle is
clearly visible from Fig. 1. The left panel shows that the maximum of Higgs line reaches
clearly out of the background while on the right panel the maximum is deeply inside the background.
Z-tagging in S3 suppresses pp ZZ and pp t tZ background. S4 is based on the equality
of the invariant masses of like-signed s or s. Fig. 2 gives a clear picture of the behavior of
signal and background for the S4 selection rule. Naturally, some freedom is needed due to the
Fig. 1. The left panel shows the distribution of events according

to scalar sum of 2 most energetic (highest pT ) positively and 2 most energetic negatively charged muons or jets ( pT ). The right panel shows the distribution of events
considering traditional pT cut for single particles. Both figures correspond to luminosity L = 30 fb1 .
Fig. 2. Distribution of events according to the ratio of reconstructed invariant masses (m ++ /m ) (no other cuts are
applied). The figure corresponds to luminosity L = 30 fb1 .
206
Table 3
Optimal pT cut for different Higgs masses and the corresponding minimal discovery luminosities: the lower (Lmin )
corresponds to the generated background in our analyses and the higher (Lmax ) corresponds to 95% upper limit of the
background error
Mass of (GeV)

Optimal pT for S2 (GeV)
MC Lmin (fb1 )
MC Lmax (fb1 )
200
300
400
500
600
700
800
900
1000
300
0.25
0.26
400
0.93
1.03
600
2.0
3.1
700
3.6
7.0
860
8
17
860
17
38
860
34
77
860
62
160
860
120
320
decay width and experimental errors of the detector. We require that the ratio m++
has to
I /mI
be in the region from 0.8 to 1.2.
While
the selection rules S1, S3 and S4 are independent of the Higgs mass, the selection rule
S2 ( pT cut) has to be optimized for a certain Higgs mass value. The cut may be set to a very
high value which eliminates all background events, but inevitable loss in signal may postpone
the discovery of new physics at LHC. Thus it is natural to take the minimal discovery luminosity
(Lmin ) as the optimization criteria. Looking for a cut value that enables to make a discovery with
the lowest luminosity, we are dealing with small signal and background expectations by definition. Simple significance estimators cannot be exploited here. We have used the log-likelihood
ratio (LLR) statistical method [44,45] to demand 5 discovery potential to be bigger than 95%
(1 CLs+b > 0.95) as for a discovery criteria. This is a rather strong requirement, because it
allows to make a discovery (meaning the fluctuation of background may mimic the outcome of
an experiment with probability less than 2.9 107 (5 )) during the specified luminosity with a
probability of 95% (if s + b hypothesis is correct). The widely used convention, that significance
should exceed five, gives only 50% discovery potential in Gaussian limit and diminishes to very
small values when background approaches zero.

The best value for S2 cut does depend on M but is not too sensitive to it. Typically the pT
can be assigned a value with a precision of 100 GeV while affecting the minimum luminosity
by only a couple of percent. In the Table 3 the approximated middle point of this value is given.
As the best S2 cut is very strong, it suppresses almost entirely the generated background (being
combined with the other selection rules). For Higgs masses above 500 GeV the background is
totally suppressed and the discovery potential criteria meets the requirement for 3 signal events
(6 invariant masses). Nevertheless we cannot infer that the background is really zero in nature. To
estimate the statistical error due to final number of generated background events we have found
95% upper limit of background according to Poisson statistics (Table 4, in brackets). Using this
limit in LLR analysis we get much higher luminosities for discovery. Even a very small background expectancy (b = 0.01) gives some possibility to have one (9.9103 ) or two (4.9105 )
background events in the experiment and these outcomes cannot be interpreted as discovery anymore. This phenomenon shifts the minimal required luminosity to much higher values denoted
as Lmax in Table 3.
An example of invariant mass distribution after applying selection rules are shown in Fig. 3
for M = 500 GeV. A tabulated example is given for M = 200, 500, 800 GeV in Table 4
corresponding to the luminosity L = 30 fb1 . The strength of S2 cut is clearly visible: almost no decrease in signal while the number of the background events descends close to final
minimum value. A peculiar behavior of S4reducing the background, while also increasing the signal in its peakis the effect of applying the correction method described
above.
207
Fig. 3. Distribution of invariant masses after applying selection rules (S1S4) for Higgs M = 500 GeV and the Standard
Model background (L = 30 fb1 ). The histogram in the right panel is a zoom of the left histogram to illustrate the effects
of the selection rules S2S4.
Table 4
Effectiveness of the selection rules for the background and signal. All event numbers in the table are normalized for
L = 30 fb1 . The numbers in brackets mark errors at 95% confidence level for Poisson statistics. The signal increases
after S4 due to the reconstructed decays
Process
N of invariant masses
N of
S1
S2
S3
S4
1534
1222 (168)
21.3 (4.0)
95.0 (12.0)
1488
172 (8.5)
15.5 (1.0)
22.5 (0.7)
1465
134 (6.9)
6.3 (1.2)
9.8 (0.5)
1539
17.6 (3.7)
2.2 (1.1)
1.7 (0.2)
48.4
178 (28)
6.6 (1.7)
9.4 (2.9)
47.5
2.1 (0.9)
2.3 (1.0)
1.4 (0.2)
46.8
1.65 (0.87)
1.0 (1.0)
0.68 (0.19)
49.5
0.10 (0.35)
0.00 (0.1)
0.08 (0.09)
5.05
77 (12)
2.6 (1.2)
2.5 (0.8)
5.00
0.00 (0.22)
0.39 (0.4)
0.34 (0.16)
4.92
0.00 (0.22)
0.39 (0.4)
0.17 (0.09)
5.21
0.00 (0.07)
0.00 (0.1)
0.00 (0.02)
Energy range 150250 GeV

M = 200 GeV
t t 4
t tZ
ZZ
4670

M = 500 GeV
t t 4
t tZ
ZZ
119.2

M = 800 GeV
t t 4
t tZ
ZZ
11.67
5. Including measurement errors to Monte Carlo

In this section we make an attempt to estimate simplified detector effects at the level of Monte
Carlo analyses. In order to do that we have added overall detection efficiencies for the Monte
Carlo generated s and -jets0.98 and 0.6, respectively. Additionally, we applied Gaussian
distortion functions to Monte Carlo produced data for s, -jets and ETmiss which were used to
alter randomly those quantities in the analysis. Although the precision of detection is sensitive
to pT of and || we use the mean values for a rough estimation. We make the following
assumptions based on [4143].
208
Table 5
Optimal
pT cuts and minimal discovery luminosities for different Higgs masses when the estimates of detector
measurement errors are taken into account. Two boundaries for the minimal luminosity are given: the lower (Lmin )
corresponds to the generated background and the higher (Lmax ) corresponds to 95% upper limit of the background error
Mass of (GeV)

Optimal pT for S2 (GeV)
Det Eff Lmin (fb1 )
Det Eff Lmax (fb1 )
200
300
400
500
600
700
800
900
1000
300
0.526
0.546
400
1.20
2.19
600
3.0
6.5
700
6.6
16.6
860
15
39
860
30
86
860
60
190
860
111
420
860
200
900
Table 6
The importance of reconstruction channels at Monte Carlo level and considering detector efficiency effects
Percentage of the channel after reconstruction
Decay channel
31j
22j
13j
correction
Monte Carlo
MC + efficiencies
21
38
28
25
26
12
9
1
16
24
The direction of muon ( -jet) is altered with the Gaussian distribution: mean = 0.0005
(0.031) and variance 2 = 0.003 (0.017). The transverse momentum is altered according to
the pT ,rec /pT ,Monte Carlo Gaussian distribution: mean = 1. (0.897) and variance 2 = 0.03
(0.089). Both components of missing transverse energy are altered independently according to
the Gaussian distribution (mean = 0 GeV and variance 2 = 25 GeV) by adding the piece to
its Monte Carlo value.
The result of such a distortion is a decrease in both signal and background approximately by
factor two (Table 5). As the background and the signal decrease proportionally, the luminosity
needed for discovery roughly doubles. Remarkably the optimized S2 cut value does not change
significantly. The proportion of the reconstruction channels in the total analysis has changed
remarkably as shown in Table 6. The reason is clearly the small detection efficiency of -jets.
The 13j channel comprises only 1% of the total signal if the detector effects are considered,
while in the pure Monte Carlo analysis it forms 9%. The additional possible detector effects
make the correction even more relevant.
6. Conclusions and outlook
We have studied possible direct test of little Higgs scenarios which light particle spectrum
includes a triplet scalar multiplet at LHC experiments. We have investigated the DrellYan pair
production of the doubly charged Higgs boson and its subsequent leptonic decays. In addition to
solving the little hierarchy problem, this scenario can also explain the origin of non-zero neutrino
masses and mixing via the triplet Higgs neutrino mass mechanism. Simple connection between
the observed neutrino mixing and triplet Yukawa couplings allows us to predict the leptonic
branching ratios of the triplet. Thus the experimental signatures of the model do not depend on
the size of the triplet Yukawa couplings allowing direct tests of this scenario at LHC.
In our analyses we have considered four and/or final states including up to 3 tau leptons.
We propose four selection rules to achieve the optimized signal and background ratio. As the
++ decays are lepton number violating, we have shown that the background can be practically
eliminated. In such an unusual situation we have used the LLR statistical method to demand
5 discovery potential to be bigger than 95% (1 CLs+b > 0.95) as the discovery criterion.
209
The results of optimized cut values are presented in Table 3. Considering the pure Monte Carlo
study, ++ up to the mass 300 GeV can be discovered in the first year of LHC (L = 1 fb1 )
and ++ up to the mass 800 GeV can be discovered for the integrated luminosity L = 30 fb1 .
Including particle reconstruction efficiencies as well as Gaussian distortion functions for the
particle momentas and missing energy which mimic detector inefficiencies at Monte Carlo level,
our results show that ++ can be discovered up to the mass 250 GeV in the first year of LHC
and 700 GeV mass is reachable for the integrated luminosity L = 30 fb1 .
For further studies of this scenario at LHC progress can be made both physics-wise as well as
technically. Full simulations of the detector effects are needed which also include the electron,
muon and tau final states. For better determination of statistical errors coming from the background studies bigger SM background datasets must be produced. This requires huge computing
resources. If these goals can be achieved, the proposed phenomenology opens a new window to
study the neutrino properties at colliders. In addition to the considerations in this paper, one can
determine at LHC experiments the hierarchy (normal or inverse) of light neutrino mass spectrum, and to estimate the two Majorana phases which are not measurable in neutrino oscillation
experiments [46].
Acknowledgements
We thank the Estonian Science Foundation for the Grant No. 6140. The work described in
this paper was also supported by the European Union through the FP6-2004-Infrastructures-6contract No. 026715 project BalticGrid.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
G.F. Giudice, C. Grojean, A. Pomarol, R. Rattazzi, hep-ph/0703164.

N. Arkani-Hamed, A.G. Cohen, H. Georgi, Phys. Rev. Lett. 86 (2001) 4757, hep-th/0104005.
H.C. Cheng, C.T. Hill, S. Pokorski, J. Wang, Phys. Rev. D 64 (2001) 065007, hep-th/0104179.
N. Arkani-Hamed, A.G. Cohen, H. Georgi, Phys. Lett. B 513 (2001) 232, hep-ph/0105239.
T. Han, H.E. Logan, B. McElrath, L.T. Wang, Phys. Rev. D 67 (2003) 095004, hep-ph/0301040.
W. Kilian, J. Reuter, Phys. Rev. D 70 (2004) 015004, hep-ph/0311095.
T. Han, H.E. Logan, L.T. Wang, JHEP 0601 (2006) 099, hep-ph/0506313.
J. Lee, hep-ph/0504136.
T. Han, H. Logan, B. Mukhopadhyaya, R. Srikanth, Phys. Rev. D 72 (2005) 053007, hep-ph/0505260.
F. del Aguila, M. Masip, J.L. Padilla, Phys. Lett. B 627 (2005) 131, hep-ph/0506063.
S.R. Choudhury, N. Gaur, A. Goyal, Phys. Rev. D 72 (2005) 097702, hep-ph/0508146.
A. Abada, G. Bhattacharyya, M. Losada, Phys. Rev. D 73 (2006) 033006, hep-ph/0511275.
J. Schechter, J.W.F. Valle, Phys. Rev. D 22 (1980) 2227.
E. Ma, U. Sarkar, Phys. Rev. Lett. 80 (1998) 5716, hep-ph/9802445.
N. Arkani-Hamed, A.G. Cohen, E. Katz, A.E. Nelson, JHEP 0207 (2002) 034, hep-ph/0206021.
J.F. Gunion, C. Loomis, K.T. Pitts, in: the Proceedings of 1996 DPF/DPB Summer Study on New Directions for
High-Energy Physics (Snowmass 96), Snowmass, Colorado, 25 June12 July 1996, pp. LTH096, hep-ph/9610237.
J.F. Gunion, J. Grifols, A. Mendez, B. Kayser, F.I. Olness, Phys. Rev. D 40 (1989) 1546.
K. Huitu, J. Maalampi, A. Pietila, M. Raidal, Nucl. Phys. B 487 (1997) 27, hep-ph/9606311.
M. Muhlleitner, M. Spira, Phys. Rev. D 68 (2003) 117701, hep-ph/0305288.
A.G. Akeroyd, M. Aoki, Phys. Rev. D 72 (2005) 035011, hep-ph/0506176.
T. Rommerskirchen, T. Hebbeker, J. Phys. G 33 (2007) N47.
S. Chang, JHEP 0312 (2003) 057, hep-ph/0306034.
C.X. Yue, S. Zhao, hep-ph/0701017.
S.R. Choudhury, A.S. Cornell, A. Deandrea, N. Gaur, A. Goyal, Phys. Rev. D 75 (2007) 055011, hep-ph/0612327.
M. Blanke, A.J. Buras, B. Duling, A. Poschenrieder, C. Tarantino, hep-ph/0702136.
210
[26] P. Minkowski, Phys. Lett. B 67 (1977) 421;

M. Gell-Mann, P. Ramond, R. Slansky, in: P. Van Nieuwenhuizen, D. Freedman (Eds.), Proceedings of the Supergravity Stony Brook Workshop, New York, 1979, North-Holland, Amsterdam;
T. Yanagida, in: A. Sawada, A. Sugamoto (Eds.), Proceedings of the Workshop on Unified Theories and Baryon
Number in the Universe, Tsukuba, Japan 1979, KEK Report No. 79-18, Tsukuba;
R. Mohapatra, G. Senjanovic, Phys. Rev. Lett. 44 (1980) 912.
[27] E. Ma, M. Raidal, U. Sarkar, Phys. Rev. Lett. 85 (2000) 3769, hep-ph/0006046.
[28] E. Ma, M. Raidal, U. Sarkar, Nucl. Phys. B 615 (2001) 313, hep-ph/0012101.
[29] E. Ma, U. Sarkar, Phys. Lett. B 638 (2006) 356, hep-ph/0602116.
[30] N. Sahu, U. Sarkar, hep-ph/0701062.
[31] G. Marandella, C. Schappacher, A. Strumia, Phys. Rev. D 72 (2005) 035014, hep-ph/0502096.
[32] G. Azuelos, et al., Eur. Phys. J. C 39S2 (2005) 13, hep-ph/0402037.
[33] T. Sjostrand, P. Eden, C. Friberg, L. Lonnblad, G. Miu, S. Mrenna, E. Norrbin, Comput. Phys. Commun. 135 (2001)
238, hep-ph/0010017.
[34] W.M. Yao, et al., Particle Data Group, J. Phys. G 33 (2006) 1.
[35] D. Acosta, et al., CMS Collaboration, The CMS Physics Technical Design Report, vol. 1, Section 10.4: Electron
reconstruction and selection, CERN/LHCC vols. 2006001, 2006. CMS TDR 8.1.
[36] D. Acosta, et al., CDF Collaboration, Phys. Rev. Lett. 93 (2004) 221802, hep-ex/0406073.
[37] V.M. Abazov, et al., D0 Collaboration, Phys. Rev. Lett. 93 (2004) 141801, hep-ex/0404015.
[38] E. Boos, et al., CompHEP Collaboration, Nucl. Instrum. Methods A 534 (2004) 250, hep-ph/0403113.
[39] A.S. Belyaev, et al., hep-ph/0101232.
[40] D. Acosta, et al., CMS Collaboration, The CMS Physics Technical Design Report, vol. 1, Section 12.2: b-tagging
tools, CERN/LHCC 2006001, 2006. CMS TDR 8.1.
[41] D. Acosta, et al., CMS Collaboration, The CMS Physics Technical Design Report, vol. 2, Physics Performance, J.
Phys. G: Nucl. Part. Phys. 34 (2007) 9951579.
[42] CMS Collaboration, The Compact Muon Solenoid. Muon Technical Design Report, CERN-LHCC-97-32.
[43] D. Acosta, et al., CMS Collaboration, The CMS Physics Technical Design Report, vol. 1, Section 11.5: Missing
transverse energy, CERN/LHCC vols. 2006001, 2006, CMS TDR 8.1.
[44] A.L. Read, Modified frequentist analysis of search results (The CL(s) method), CERN-OPEN-2000-205.
[45] R. Barate, et al., LEP Working Group for Higgs boson searches, Phys. Lett. B 565 (2003) 61, hep-ex/0306033.
[46] M. Kadastik, M. Raidal, L. Rebane, in preparation.
Nuclear Physics B 787 [FS] (2007) 211240
BRST approach to Lagrangian construction

for fermionic higher spin fields in AdS space
I.L. Buchbinder a , V.A. Krykhtin a,b, , A.A. Reshetnyak c
a Department of Theoretical Physics, Tomsk State Pedagogical University, Tomsk 634041, Russia
b Laboratory of Mathematical Physics, Tomsk Polytechnic University, Tomsk 634034, Russia
c Laboratory of Non-equilibrium State Theory, Institute of Strength Physics and Materials Science,
Tomsk 634021, Russia

Received 25 April 2007; received in revised form 8 June 2007; accepted 11 June 2007
Abstract
We develop a general gauge-invariant Lagrangian construction for half-integer higher spin fields in the
AdS space of any dimension. Starting with a formulation in terms of an auxiliary Fock space, we obtain closed nonlinear symmetry algebras of higher spin fermionic fields in the AdS space and find the
corresponding BRST operator. A universal procedure for constructing gauge-invariant Lagrangians describing the dynamics of fermionic fields of any spin is developed. No off-shell constraints for the fields
and gauge parameters are imposed from the very beginning. It is shown that all the constraints determining an irreducible representation of the AdS group arise as a consequence of the equations of motion and
gauge transformations. As an example of the general procedure, we derive gauge-invariant Lagrangians
for massive fermionic fields of spin 1/2 and 3/2 containing the complete set of auxiliary fields and gauge
symmetries.
1. Introduction
Problems of higher spin field theory have attracted much attention for a long time due to the
hope of finding new possibilities and approaches to the unification of the fundamental interac* Corresponding author at: Laboratory of Mathematical Physics, Tomsk Polytechnic University, Tomsk 634034,
Russia.
E-mail addresses: joseph@tspu.edu.ru (I.L. Buchbinder), krykhtin@mph.phtd.tpu.edu.ru (V.A. Krykhtin),
reshet@tspu.edu.ru (A.A. Reshetnyak).
doi:10.1016/j.nuclphysb.2007.06.006
212
I.L. Buchbinder et al. / Nuclear Physics B 787 [FS] (2007) 211240
tions. Higher spin field theory is closely related to superstring theory, which operates with an
infinite tower of bosonic and fermionic higher spin fields, including massless and massive ones.
A Lagrangian formulation for an interacting higher spin field theory is one (perhaps the last) of
the unsolved general problems of classical field theory (see, e.g., the reviews [1]).
The current studies in higher spin field theory concern the aspects of Lagrangian formulation
in various dimensions, searches for supersymmetric generalizations, finding a correspondence
with superstring and M theories and a construction of interacting Lagrangians (see [29] for
massive and [1018] for massless higher spin theories).
It is well known that a space of constant curvature, in particular, the AdS space, is the simplest
non-trivial background providing a consistent propagation of higher spin fields (see, e.g., [2,19]).
We point out two attractive features of higher spin theory in the AdS space. First, the radius of
the AdS space ensures the presence of a natural dimensional parameter for the accommodation of
compatible self-interactions for massless higher spin fields [20,21]. Second, higher spin fields in
the AdS space are closely related to the conjecture [22] on the AdS/CFT correspondence between
the conformal N = 4 SYM theory and the superstring theory on the AdS5 S5 RamondRamond
background [7].
The study of a Lagrangian formulation for massive higher spin field theories was initiated by
the pioneering works of FierzPauli and SinghHagen [23] for theories in the four-dimensional
Minkowski space. These works demonstrated a specific feature of higher spin field theories: all
such theories include, apart from the basic fields with a given spin, also some auxiliary fields with
lesser spins, which provide a compatibility of the Lagrangian equations of motion with the constraints determining an irreducible representation of the Poincare group. Attempts to construct
Lagrangian descriptions of free higher spin field theories in curved spaces and derive higher
spin interactions have resulted in consistency problems which generally remain open in spite of
numerous efforts.
The present work is devoted to a derivation of general gauge-invariant Lagrangian for massive
and massless fields of any half-integer spin in AdS spaces of arbitrary dimensions. Our approach
is based on use of the BRSTBFV construction [24] (see also the reviews [25]), which was initially developed for the quantization of gauge theories. Following a tradition accepted in string
theory and in higher spin field theory, we further refer to this construction as the BRST construction, and to the corresponding BFV charge, as the BRST charge. The general application
of the BRST construction to higher spin field theory consists in three steps. First, the conditions determining the representation with a given spin are regarded as first-class constraints.
Second, using the algebra of these constraints, one constructs the BRST charge. Third, a higher
spin field Lagrangian is constructed in terms of the BRST charge in such a way that the corresponding equations of motion reproduce the initial constraints. The final step is of fundamental
importance and ensures the correctness of the Lagrangian construction. We emphasize that this
approach automatically implies a gauge-invariant Lagrangian. One can expect that in the case of
massive theories a Lagrangian resulting from the BRST construction should contain a number of
Stuckelberg fields.
A derivation of Lagrangians for massless bosonic higher spin fields on a basis of the BRST
approach has been considered in [26,27] for the flat space and in [28] for the AdS space. A generalization of the BRST approach to Lagrangians of massless fermionic higher spin fields in flat
spaces of arbitrary dimensions has been given in [29]. An interaction vertex for massless bosonic
higher spin fields in flat and AdS spaces has been studied within the BRST approach in [30].
The first implementation of the BRST construction for the derivation of Lagrangians for massive
higher spin fields has been given in our papers for bosonic [31] and fermionic [32] models in
213
Minkowski spaces of arbitrary dimensions and for bosonic models in AdS spaces of arbitrary dimensions [33]. Thus, the BRST approach presents a universal generating method of constructing
Lagrangian formulations for higher spin fields.
In this paper, we develop a gauge-invariant approach to a Lagrangian construction for totally
symmetric fermionic higher spin fields in the AdS space of any dimension.1 The conditions
determining an irreducible representation of the AdS group with a given spin on tensorspinor
fields are reformulated as operator constraints in an auxiliary Fock space. These constraints form
a nonlinear superalgebra with a central bosonic charge. Such an algebra has a more complex
structure in comparison with its counterpart for bosonic higher spin fields in the AdS space [33],
due to the presence of non-vanishing operator structure functions resolving the Jacobi identity,
and therefore the construction of the corresponding BRST charge faces a problem. Some aspects
of the BRST construction for bosonic nonlinear constraint algebras have been discussed in [35,
36]. We find a solution of the BRST construction for a nonlinear superalgebra in the theory under
consideration and obtain a Lagrangian which reproduces the initial constraints as a consequence
of the equations of motion.
The paper is organized as follows. In Section 2, we examine a closed operator superalgebra
based on the constraints that determine an irreducible representation of the AdS group with a
half-integer spin. In Section 3, we find a superalgebra of the additional parts for the superalgebra of a modified set of the initial operators, obtained by a linear transformation of the initial
constraints, and then realize its representation in terms of new (extra) creation and annihilation
operators. In Section 4, we calculate a deformed nonlinear superalgebra of constraints enlarged
by the additional parts of the modified constraints. A similar construction for fermionic fields
in flat space has been given in [29]. Next, we construct a BRST operator corresponding to the
superalgebra of the modified enlarged constraints. The derivation of an action and of a sequence
of reducible gauge transformations describing the propagation of a fermionic field of an arbitrary
spin in the AdS space is realized in Section 5. In Section 6, we prove that the constructed action
reproduces the correct conditions for the field that determine an irreducible representation of the
AdS group with a fixed half-integer spin. In Section 7, we illustrate the procedure by constructing
gauge-invariant Lagrangians for the fields of spin 1/2 and 3/2. In conclusion, we summarize the
results of the work and discuss some open problems.
In addition to the conventions of Refs. [32,33], we use the notation (A), gh(A) for the respective values of Grassmann parity and ghost number of a quantity A, and denote by [A, B} the
supercommutator of quantities A, B, which in the case of definite values of Grassmann parity is
given by [A, B} = AB (1)(A)(B) BA.
2. Auxiliary Fock space for higher spin fields in AdS spacetime
The massive half-integer spin s = n + 12 representations of the AdS group are realized in the
space of totally symmetric tensorspinor fields 1 ...n (x), the Dirac index being suppressed,
satisfying to the conditions (see, e.g., [4])

1
d
i r 2 n + 2 m 1 2 n (x) = 0,
(1)
2
2 n (x) = 0,
(2)
1 For the sake of completeness, we note that a Lagrangian formulation for massive fields of an arbitrary half-integer
spin in the AdS space subject to algebraic constraints has been recently suggested in [34].
214
R
where r = d(d1)
, with R being the scalar curvature of the spacetime. We use the metric with
the mostly minus signature, and the Diracs matrices obey the relation
{ , } = 2g .
(3)
In order to avoid an explicit manipulation with the indices, it is convenient to introduce an

auxiliary Fock space2 H generated by creation and annihilation operators with tangent space
indices (a, b = 0, 1, . . . , d 1)

aa , ab+ = ab , ab = diag(+, , . . . , ).
(4)
An arbitrary vector in this Fock space has the form
| =
a1 an (x)a +a1 a +an |0 =
1 n (x)a +1 a +n |0,
n=0
+
+a
a
a (x) = ea (x)a , a (x) = ea (x)a ; ea (x)
(5)
n=0
where

a , a+ = g .
being the vielbein. It is evident that

(6)
We refer to the vector (5) as the basic vector. The fields 1 n (x) are the coefficient functions
of the vector | and its symmetry properties are stipulated by the symmetry properties of the
product of the creation operators. We also assume the standard relation
a
e + a b eb = 0.
ea = ea
(7)
We intend to realize relations (1), (2) as certain constraints on the vector | (5). To this end,
we define an operator D acting on the vectors |,

1
1
ab
+
D = aa ab ab ,
(8)
|0 = 0, ab = (a b b a ),
4
2
and operators
1
t0 = i D m r 2 (g0 2),
t1 = a ,
d
g0 = a + a + .
2
We can see that the constraints
t0 | = t1 | = 0
(9)
(10)
(11)
(12)
for the basic vector (5) are equivalent to Eqs. (1), (2), with each component 1 n (x) in (5)
subject to (1), (2), and therefore they describe a field of spin n + 1/2.
Because of the fermionic nature of Eqs. (1), (2) with respect to the standard Grassmann parity,
and due to the bosonic nature of the operators t0 , t1 : (t0 ) = (t1 ) = 0, in order to equivalently
transform these operators into fermionic ones, we now introduce a set of d + 1 Grassmann-odd
gamma-matrix-like objects , subject to3

, = 2g ,
, = 0,
2 = 1
(13)
2 For more details, see [33].
3 For more details, see [32].
215
and related to the conventional gamma-matrices as follows:

= .
(14)
In terms of these gamma-matrices, we define the Grassmann-odd constraints

1
t0 = i D + m + r 2 (g0 2) ,
t1 = a ,
(15)
related to (9), (10) as follows:

t0 = t0 ,
t1 = t1 .
(16)
In order to obtain a Hermitian BRST operator, being the central object of the Lagrangian construction in the BRST approach, we need a set of first-class constraints which must be invariant
with respect to Hermitian conjugation and must form a superalgebra with respect to supercommutator multiplication. To this end, we define an odd scalar product:

0|a 1 a k +1 k (x)0 1 n (x) a +1 a +n |0.
| = d d x |g|
(17)
n,k=0
The operators t0 , t1 (15) and t1+ = a+ , being the Hermitian conjugate of t1 with respect to the
1
scalar product (17), generate an operator superalgebra with the central charge m
= (m 2r 2 )
and the following set of operators:

1
t0 = i D + m + r 2 (g0 2) ,
(18)
t1 = a ,
t1+ = a+ ,
(19)
l1 = ia D ,
1
l 2 = a a ,
2
l1+
(20)
g0 = a+ a +
= ia D ,
1
l2+ = a + a+ ,
2
d
,
2
(21)

2

1
d(d 3)
D r g0 + t1+ t1 +
+ m + r 2 (g0 2) ,
l0 = g D D
4
(22)
(23)
which is invariant under Hermitian conjugation. The operators (18)(23) form an algebra given
by Table 1, where
+
1
t0 , t1 = 2l1+ + r 2 t1+ ,
(24)

1
t0 , t1 = 2l1 r 2 t1 ,
(25)
+
+

1

t1 , l1 = l1 , t1 = t0 m
(26)
+ r 2 g0 ,

1
1
t0 , l1 = r 2t1+ l2 + g0 t1 t1 r 2 l1 ,
(27)
2

+
1
1
l1 , t0 = r 2l2+ t1 + t1+ g0 t1+ r 2 l1+ ,
(28)
2

1
1,
t1 , l0 = r 4t1+ l2 + 2g0 t1 t1 r 2 mt
(29)
+
+

1
l0 , t1 = 4r l2 t1 + t1+ g0 + 2r 2 mt
(30)
1+ ,
216
Table 1
The algebra of the initial operators
t0
[, }
t0
t1
2l0
t1
t1+
l
0
l1
l1+
l0
l1
l1+
(25)
(24)
(27)
(28)
2 r
4l2
2g0
2g0
4l2+
(29)
(30)
0
(26)
(26)
0
0
t1
0
(27)
(28)
(29)
0
(26)
(30)
(26)
0
0
(31)
(32)
(31)
0
(35)
(32)
(35)
0
l2+
2 r
g0
m
0
0
1 +
2l
0
t1+
t1
0
t1
(33)
(34)
t1+
0
0
0
0
l1+
l1
0
l1
l2+
l2
(25)
(24)
2 r 2 l2
l2
t1+
1
2 l2
2 r
t1+
0
1 +
2l
g0
t1
t1+
0
0
0
l1
l1+
0
0
0
0
(33)
0
l1
(34)
l1+
0
g0
2l2
2l2+
0
0
0
0
g0
l1+
0
2l2
0
2l2+
0

1
l1 , l0 = 4r l1+ l2 + g0 l1 + 2r 2 ml
1,
+
+

1
l0 , l1 = 4r l2 l1 + l1+ g0 + 2r 2 ml
1+ ,

1 1
l2 , l0 = 4r 2 r 2 (1 + g0 ) + m
l2 ,
+
1

1
l0 , l2 = 4r 2 l2+ r 2 (1 + g0 ) + m
,

1
3
1
2 2r 2 mg
0 + r t1+ t1 g0 4l2+ l2 .
l1 , l1+ = l0 m
2
2

(31)
(32)
(33)
(34)
(35)
We call this algebra the massive half-integer higher spin symmetry algebra in the AdS space.
The method of Lagrangian construction requires an enlarging of the initial constraints oi
(t0 , t1 {oi }), so that the enlarged Hermitian operators contain arbitrary parameters and the set
of enlarged operators form an (super)algebra. A procedure of constructing of these enlarged constraints Oi = oi + oi for the operators oi is considerably simplified if the initial operators oi
(super)commute with its additional parts oi : [oi , oj } = 0, defined in the auxiliary operator space.
In this case we can apply the method elaborated in [33]. The such realization of the enlarged
first-class constraints Oi , by means of an additive composition of the quantities oi with additional parts oi , requires a transformation of oi into an equivalent set of constraints o i which do
not contain the -matrix, because the set oi will contain this object by construction. Thus, in
order to simplify the subsequent calculations, we choose the operators o i related by a nondegenerate linear transformation4 oi = Uji o j with the initial constraints reproducing relations (1), (2),
t0 = i D ,

d(d 3)
D r g0 + t1+ t1 +
.
l0 = g D D
4
(36)
(37)
The other constraints coincide with the initial ones.

4 Note that the case of a two-parametric set of constraints determined by the operator t = i D + m
1 +
0
1
2 r 2 g0 , analogous to the consideration for bosonic fields in the AdS space [33], from which it would be possible to
construct a Lagrangian that reproduces Eq. (12) for any values 1 , 2 , can be realized by means of the above nondegenerate transformation of the superalgebra of the initial constraints.
217
Table 2
The algebra of the modified initial operators
[, }
t0
t0
t1
t1+
l0
l1
l1+
l2
l2+
2l0
2l1
2l1+
0
(38)
(41)
0
0
g0
m
0
0
t1
2l1
4l2
2g0
(39)
0
t0
0
t1+
t1
0
t1+
l0
l1
l1+
l2
l2+
g0
(42)
t0
0
t1
0
0
(39)
(42)
0
(40)
(43)
0
0
(41)
t0
0
(43)
(44)
0
l1
0
0
0
t1
0
0
l1
0
g0
0
t1+
0
0
l1+
0
g0
0
0
0
l1+
0
0
t1
t1+
0
l1
l1+
2l2
2l2+
0
0
0
0
0
0
0
0
t1+
0
(38)
0
t0
(40)
0
(44)
0
l1+
2l2
0
0
0
0
0
2l1+
2g0
4l2+
l1
0
2l2+
0
The operators o i given by (19)(22), (36), (37) with the central charge m
form a superalgebra
given by Table 2, where

1
[t0 , l1 } = r 2t1+ l2 + g0 t1 t1 ,
(38)
2

[t1 , l0 } = r 4t1+ l2 + 2g0 t1 t1 ,
(39)

+
[l1 , l0 } = r 4l1 l2 + 2g0 l1 l1 ,
(40)

+
1
l1 , t0 = r 2l2+ t1 + t1+ g0 t1+ ,
(41)
2

l0 , t1+ = r 4l2+ t1 + 2t1+ g0 t1+ ,
(42)

+
+
+
+
l0 , l1 = r 4l2 l1 + 2l1 g0 l1 ,
(43)

1
3
l1 , l1+ = l0 + r g02 g0 4l2+ l2 + t1+ t1 .
(44)
2
2
In terms of the operators o i , equations (12) have the form

1
t0 + m + r 2 (g0 2) | = 0,
t1 | = 0.
(45)
In what follows, we shall demonstrate the construction of Lagrangians by using the BRST
approach developed for bosonic fields in the AdS space in [33] so as to reproduce equations (12),
or, equivalently, (45). According to our approach, we have to extend the operators o i of the
algebra given by Table 2 by additional parts oi : o i O i = o i + oi , so that (1) O i must be in
involution [O i , O j } O k and (2) each Hermitian operator must contain linearly an arbitrary
parameter whose values are to be determined later (for details, see [29,32,33]).
The next step of this procedure is to find the additional parts oi for the operators (19)(22),
(36), (37).
3. Additional parts of operators
As stated at the end of the previous section, now we intend to find the additional parts oi
for the operators (19)(22), (36), (37). Following the procedure described in the bosonic case
in [33], we must first find the superalgebra of the additional parts. Then these additional parts
are constructed from new (additional) creation and annihilation operators, as well as from the
218
constants of the theory r, m, and from one of the gamma-matrix-like objects . Furthermore, we
must introduce linearly5 into t0 , g0 (these are the additional parts for t0 , g0 , respectively) some
arbitrary constants whose values will be determined later by the condition that equations (45) be
reproduced.
Let us briefly remind and at the same time generalize the method given in [33] to algebras including fermionic operators. The (super)commutation relations for the operators o i from Table 2
have the structure
[o i , o j }s = fijk o k + fijkm o k o m ,
(46)
where the constants fijk , fijkm obey the properties (fijk , fijkm ) = (1)(Oi )(Oj ) (fjki , fjkm
i ). Then
we suppose that the operators o i supercommute with the additional parts oi . In this case, one can
check that if we define the superalgebra of the additional parts in the form

oi , oj s = fijk ok (1)(ok )(om ) fijkm om
(47)
ok ,
then the enlarged operators O i = o i + oi will form a closed superalgebra

[O i , O j }s = fijk O k fijmk + (1)(Ok )(Om ) fijkm om
O k + fijkm O k O m ,
(48)
which is deformed in comparison with (46).

After applying of the procedure above, the superalgebra of the additional parts takes the form
given by Table 3, where

1
+

l1 , t0 = r 2t1 l2 + g0 t1 t1 ,
(49)
2

+
1
t0 , l1 = r 2l2+ t1 + t1+ g0 t1+ ,
(50)
2
+

l0 , t1 = r 4t1 l2 + 2g0 t1 t1 ,

(51)
+
+

t1 , l0 = r 4l2 t1 + 2t1+ g0 t1+ ,
(52)

+

l0 , l1 = r 4l1 l2 + 2g0 l1 l1 ,
(53)
+
+
+
+
l1 , l0 = r 4l2 l1 + 2l1 g0 l1 ,
(54)

+
1
3
l1 , l1 = l0 r g02 g0 4l2+ l2 + t1+ t1 .
(55)
2
2
In accordance with our method, we ascribe the following value to the additional central charge:
so that the enlarged central charge M becomes equal to zero.
m
= m,
Explicit expressions for the additional parts can be found by the method described in the
papers [29,36] and extended to the case of the Verma module construction for a nonlinear superalgebra given by Table 3. Omitting tedious calculations, we have
t1+ = f + + 2b2+ f,
g0
= b1+ b1
+ 2b2+ b2
+ f f + h,
l1+ = m1 b1+ ,
(56)
l2+
(57)
= b2+ ,
5 We have to introduce linearly some arbitrary constants into all of the additional parts corresponding to Hermitian
operators. Since t02 = l0 , we cannot obtain an independent arbitrary constant in l0 .
219
Table 3
The algebra of the additional parts for the operators
[, }
t0
t1
t1+
l0
t0
t1
t1+
l0
l1
l1+
l2
l2+
g0
2l0
2l1
2l1+
2l1
4l2
2g0
2l1+
2g0
4l2+
(49)
(51)
(51)
(52)
(49)
t0
(53)
(50)
0
t0
0
t1
(54)
0
0
t1+
t0
= 2m1 b1+ f
(55)
0
l1+
(50)
g0
t1+
0
(54)
t1
0
(55)
0
l1
0
l1+
l1
l2+
l2
t0
t1+
t1
t0
(53)
(52)
0
l1+
l1
l1
0
g0
2l2
t1
0
0
t1+
0
l1+
l1
0
g0
l1+
2l2
2l2+
2l2+

m1 +
2r k (b2+ )k1 b12k
+ +
f 2b2 f b1
2
(2k)!
m21
k=1

r(h 12 ) +
2r k (b2+ )k b12k+1
f 2b2+ f
,
(2k)!
m1
(2k + 1)!
m21
k=0
k=0
(58)

2r k (b2+ )k1 b12k
1
1 +
h
f 2b2+ f
t1 = 2g0 f f + 2b2+ f b2 +
2
2
(2k)!
m21
k=1
+ m0

2r
m21
k
(b2+ )k b12k

1 +
2r k (b2+ )k1 b12k+1
f 2b2+ f b1+
2
(2k + 1)!
m21
k=1
m0
m1
l0 = m20 r

k=0
2r
m21
k
(b2+ )k b12k+1
,
(2k + 1)!
(59)

m0 +
8r k (b2+ )k b12k+1
f 2b2+ f
1 4k
2
m1
(2k + 1)!
m1
k=1
rb1+

8r k (b+ )k b2k+1
2
k=0
m21
(2k + 1)!
2h 4

m0 2r k (b2+ )k+1 b12k+1
+ 4r
f
m1
(2k + 1)!
m21
k=0

+ 2
1 2r k+1 (b2+ )k+1 b12k+2
8r k (b2+ )k b12k+2
2r
b
+r h
1
2
(2k + 2)!
(2k + 2)!
m21
m21
k=0
k=0

+ k 2k
b1+ b1
2r k (h 12 )
2rf f
+
b 2 b1
2
(2k)!
(2k + 1)!
m1
+
k=0

m20 r(h2 14 )
8r k+1 (b2+ )k+1 b12k+2
,
2
(2k + 2)!
m21
k=0
(60)
220
l1

2b1+ b1 + k1 2k
m1 + 8r k 2h 4k
+
b1
b
+
b2
4 1
(2k)!
(2k + 1)!
m21
k=1

m0 +
8r k (b2+ )k1 b12k
+
f 2b2 f
1 4k
2
4
(2k)!
m1
k=1

r(h 12 ) 2r k (b2+ )k b12k+1 m1 + +
2r k (b2+ )k1 b12k
f
f
+
b
2m1
(2k + 1)!
2 1
(2k)!
m21
m21
k=0
k=1

r(h 12 ) +
2r k (b2+ )k b12k+1
2r k (b2+ )k b12k
f f
f
m
0
m1
(2k + 1)!
(2k)!
m21
m21
k=0
k=0

m20 r(h2 14 )
8r k (b2+ )k b12k+1
,
m1
(2k + 1)!
m21
= m1 b1+ b2
+
+
(61)
k=0

m20 r(h2 14 )
8r k (b2+ )k b12k+2
(2k + 2)!
m21
m21
k=0

r(h 12 )
2r k (b2+ )k b12k+2 m0 2r k (b2+ )k b12k+1
f
+
(2k + 2)!
m1
(2k + 1)!
2m21 k=0 m21
m21
k=0

8r k (b2+ )k1 b12k+1
m0 +
f 2b2+ f
1 4k
2
4m1
(2k + 1)!
m1
k=1

2b1+ b1 + k1 2k+1
1 + 8r k 2h 4k
b1
b1
+
b2
4
(2k + 1)! (2k + 2)!
m21
k=1

+ k1 2k
b1+ b1
1 + 2r k h 12
b1 .
f f
+
b2
2
2
(2k)!
(2k + 1)!
m1
l2 = g0 b2 b2+ b22
(62)
k=1
In the above expressions, h is an arbitrary dimensionless constant, while m0 and m1 are arbitrary
constants with the dimension of mass. To obtain explicit expressions for the additional parts
(56)(62) in the Fock space, we have introduced two new pairs of bosonic and a new pair of
fermionic creation and annihilation operators satisfying the standard commutation relations

b2 , b2+ = 1,
f, f + = 1.
b1 , b1+ = 1,
(63)
The resulting additional parts of the operators possess all the necessary properties, in particular,
the additional parts which correspond to Hermitian operators contain arbitrary parameters (see
the footnote 5): the operator t0 contains the arbitrary parameter m0 ; the operator g0 contains the
arbitrary parameter h. The values of the arbitrary parameters h and m0 will be determined later
by the condition that the correct equations of motion (12), or, equivalently, (45), be reproduced.
The massive parameter m1 remains arbitrary and can be expressed in terms of other parameters of the theory,
m1 = f (m, r) = 0.
This arbitrariness does not affect the equations for the basic vector (5).
(64)
221
Note that the additional parts do not obey the usual properties
+
l0 = l0 ,
+
t0 = t0 ,
+
l1 = l1+ ,
+
t1 = t1+ ,
+
l2 = l2+ ,
(65)
(66)
if one should use the standard rules of Hermitian conjugation for the new creation and annihilation operators,
(b1 )+ = b1+ ,
(b2 )+ = b2+ ,
(f )+ = f + .
(67)
To restore the proper Hermitian conjugation properties for the additional parts, we change the
scalar product in the Fock space H generated by the new creation and annihilation operators as
follows:
1 |2 new = 1 |K |2 ,
(68)
for any vectors |1 , |2 with some, yet unknown, operator K . This operator is determined by
the condition that all of the operators of the algebra must have the proper Hermitian properties
with respect to the new scalar product:
1 |K l0 |2 = 2 |K l0 |1 ,
1 |K t0 |2 = 2 |K t0 |1 ,
(69)
1 |K l1 |2 = 2 |K l1+ |1 ,
1 |K l2 |2 = 2 |K l2+ |1 ,
1 |K t1 |2 = 2 |K t1+ |1 ,
1 |K g0 |2 = 2 |K g0 |1 .
(70)
(71)
These relations permit one to determine the operator K as follows:

K = Z + Z,
Z=
1 + n1

l
1
n1 ,n2 =0 s=0
m1
t1+
l2+
n2
|0V
1
0|b1n1 f s b2n2 ,
n1 !n2 !
(72)
where the auxiliary vector |0V obeys the relations

t1 |0V = l1 |0V = l2 |0V = 0,
t0 |0V
l0 |0V
= m0 |0V ,
V 0|0V
= m20 |0V ,
= 1,
g0 |0V
(73)
= h|0V .
(74)
For low numbers n1 + nf + 2n2 , where ni are the numbers of particles associated with bi+ ,
and nf is the number of particles associated with f + , the operator K reads
K = |00| +
+
m20 rh(h 12 )
m21
b1+ |00|b1 2hf + |00|f

m0 +
f |00|b1 + b1+ |00|f + .
m1
(75)
This expression for the operator K will be used later in constructing of examples in Section 7.
Thus, in this section, we have constructed the additional parts (56)(62) for the constraints,
which obey all the requirements. In the next section, we determine the algebra of the enlarged
constraints and construct a BRST operator corresponding to this algebra.
222
Table 4
The algebra of the enlarged operators
[, }
T0
T0
T1
T1+
L0
L1
L+
1
L2
L+
2
2L0
2L1
2L+
1
0
(76)
(77)
0
0
G0
T1
2L1
4L2
2G0
(78)
0
T0
0
T1+
T1
T1+
L0
L1
L+
1
L2
L+
2
G0
(79)
T0
0
T1
0
0
(78)
(79)
0
(80)
(81)
0
0
(76)
0
T0
(80)
0
(82)
0
L+
1
(77)
T0
0
(81)
(82)
0
L1
0
0
0
T1
0
0
L1
0
G0
0
T1+
0
0
L+
1
0
G0
0
0
T1
T1+
0
L1
L+
1
2L2
2L+
2
T1+
L1
L+
1
2L2
2L+
1
2G0
4L+
2
2L+
2
4. Deformed algebra and BRST operator

Let us turn to the algebra of the enlarged operators O i = o i + oi . Since the algebra is quadratic,
there exist different possibilities of operator ordering in the right-hand side of the commutation
relations. Analogous situation takes place in the bosonic higher spin theory as well. In [33]
we have studied dependence of BRST construction on ordering prescription for the quadratic
constraint algebra and proved that different ordering prescriptions lead to equivalent Lagrangians
and the same solution to the equations of motion. Since the constraint algebra in the fermionic
case is analogous to the bosonic one, the same consequence about the final Lagrangians and the
equations of motion will be valid also for the fermionic case. Therefore we are not going to study
here all the possibilities of ordering prescriptions as this was done in [33], but choose only one
of them, which corresponds to the supersymmetrized ordering for the constraints and lead to the
expression for the BRST operator which has non-vanishing terms of the third degree in powers
of ghosts.6 The algebra of the enlarged operators corresponding to this BRST operator is given
by Table 4, where

+
1
1
+
+

+
[T0 , L1 } = r G0 T1 + T1 G0 + T1 L2 + L2 T1 g0 T1 t1 G0 2 t1 L2 + l2 T1 ,
2
2
(76)

+
1 +
1
+
+
+
L1 , T0 = r T1 G0 + G0 T1+ + L+
2 T1 + T1 L2 t1 G0 g0 T1
2
2

+
+
2 l2 T1 + t1 L2 ,
(77)

[T1 , L0 } = r G0 T1 + T1 G0 + 2T1+ L2 + 2L2 T1+ 2 g0 T1 + t1 G0 4 t1+ L2 + l2 T1+ ,
(78)

+
+
+
L0 , T1+ = r T1+ G0 + G0 T1+ + 2L+
T
+
2T
L
2
t
G
+
g
T
1
1 2
0
0 1
1

2
(79)
4 l2+ T1 + t1 L+
,
2
6 Using our results in the bosonic case [33], we see that the BRST operators corresponding to different choices of the
mentioned ordering of constraints will lead to equivalent Lagrangian formulations. At the same time, we note that for the
purpose of constructing Lagrangians it is sufficient to find only one BRST operator with a some fixed ordering for the
constraints in its commutation relations.
223

+

+

+
[L1 , L0 } = r G0 L1 + L1 G0 + 2L+
1 L2 + 2L2 L1 2 l1 G0 + g0 L1 4 l1 L2 + l2 L1 ,
(80)

+

+
+
+
+
+
L0 , L+
=
r
L
G
+
G
L
+
2L
L
+
2L
L
2
l
G
+
g
L
0
0
1
1
0
0 1
1
1
1
2
2
1

+
+
(81)
4 l1 L2 + l2 L1 ,

3 +
3
+
+
+
2
L1 , L+
1 = L0 + r G0 2L2 L2 2L2 L2 + 4 T1 T1 4 T1 T1

3 +

+

+
+
(82)
+ r 2g0 G0 + 4 l2 L2 + l2 L2 t1 T1 t1 T1 .
2
The construction of a nilpotent fermionic BRST operator for a nonlinear superalgebra is based
on the same principles as those developed in [28,33] (for a general consideration of operator BFV
quantization, see the reviews [25]). The BRST operator constructed on a basis of the algebra
given by Table 4 can be calculated with the help of the (CP)-ordering of the ghost coordinate C i
and momenta Pi operators:
+
+
Q = q0 T0 + q1+ T1 + q1 T1+ + 0 L0 + 1+ L1 + 1 L+
1 + 2 L2 + 2 L2 + G G0

+ i 1+ q1 1 q1+ p0 i G q1 + 2 q1+ p1+ + i G q1+ + 2+ q1 p1

+ q02 1+ 1 P0 + 2q1 q1+ 2+ 2 PG + G 1+ + 2+ 1 2q0 q1+ P1

+ 1 G + 1+ 2 2q0 q1 P1+ + 2 G 2+ q1+2 P2 + 2 2 G q12 P2+

i

1
+ r q0 1+ + 2q1+ 0
G0 p1 + T1 PG + T1+ P2 + iL2 p1+ t1 PG ig0 p1
2
2

2 il2 p1+ + t1+ P2

i
1
+
+
+ r(q0 1 + 2q1 0 ) G0 p1+ T1+ PG T1 P2+ iL+
2 p1 + t1 PG + ig0 p1
2
2

+ 2 il2+ p1 + t1 P2+
1
1
+

G0 P1 + L1 PG + L+
1 P2 + L2 P1 l1 PG g0 P1
2
2

2 l1+ P2 + l2 P1+
+ 2r0 1+
1
1
+
+
PG + L1 P2+ + L+
G0 P1+ + L+
2 P1 l1 PG g0 P1
2
2 1

+

+
2 l2 P1 + l1 P2
2r0 1

3i +
3i
+
+
r1 1+ 2L+
P
+
2L
P
G
P
p
+
p
T
T
2 2
0 G
1
1 1
2 2
4 1
4

3i +
+ +
+

+
+ r1 1 4 l2 P2 + l2 P2 2g0 PG
t p1 t1 p1
2 1
224

+ 2 i + +

2
+ T1 p1 P2 + T1 p1 P2+
G0 p1+ p1 + 2P2+ P2 + L+
2 p 1 + L2 p 1
2

i +
+ T1 p1 + T1 p1+ PG 2 L2 P2+ L+
(83)
2 P2 PG .
4
+r
0 1 1+
Here, q0 , q1 , q1+ and 0 , 1+ , 1 , 2+ , 2 , G are, respectively, the bosonic and fermionic ghost
coordinates corresponding to their canonically conjugate ghost momenta p0 , p1+ , p1 , P0 ,
P1 , P1+ , P2 , P2+ , PG . They obey the (anti)commutation relations

1 , P1+ = P1 , 1+ = 2 , P2+ = P2 , 2+ = {0 , P0 } = {G , PG } = 1,

[q0 , p0 ] = q1 , p1+ = q1+ , p1 = i
(84)
and possess the standard ghost number distribution, gh(C i ) = gh(Pi ) = 1, providing the property gh(Q ) = 1.
The resulting BRST operator Q is Hermitian with respect to the scalar product | in the
tensor product of the Fock spaces Htot = H H Hgh , which is constructed as the direct
product of the scalar products on H, H and Hgh ,
1 |Q+ K|2 = 1 |KQ |2
with the operator K defined in Htot and being the tensor product of the operator
1 gh ,
operators 1,
K = K 1 1 gh .
(85)
K
and the unit

(86)
Thus, we have constructed the Hermitian BRST operator. In the next section, this operator is
applied to construct Lagrangians of fermionic higher spin fields in the AdS space.
5. Construction of Lagrangians
In this section, we construct Lagrangians of fermionic massive higher spin fields in the AdS
space. This construction goes along the lines of [29,32]. We should first extract the dependence
of the BRST operator Q (83) on the ghosts G , PG ,
Q = Q + G ( + h) + BP G ,
(87)
where
+
+
Q = q0 T0 + q1+ T1 + q1 T1+ + 0 L0 + 1+ L1 + 1 L+
1 + 2 L 2 + 2 L 2

+

+ i 1 q1 1 q1+ p0 i2 q1+ p1+ + i2+ q1 p1 + q02 1+ 1 P0

+ 2+ 1 2q0 q1+ P1 + 1+ 2 2q0 q1 P1+ 2q1+2 P2 2q12 P2+

+
+
+ i
+
+
+

+ r q0 1 + 2q1 0
G0 p1 + T1 P2 + iL2 p1 ig0 p1 2 il2 p1 + t1 P2
2

+

i
+
+
+
+
+
+ r(q0 1 + 2q1 0 ) G0 p1 T1 P2 iL2 p1 + ig0 p1 + 2 il2 p1 + t1 P2
2

+
1
+
+
+

+
+ 2r0 1
G0 P1 + L1 P2 + L2 P1 g0 P1 2 l1 P2 + l2 P1
2

+
1
+
+
+
+
+
2r0 1 G0 P1 + L1 P2 + L2 P1 g0 P1 2 l2 P1 + l1 P2
2

3i +
3i
+
+
+
2L2 P2 + 2L2 P2 T1 p1 + T1 p1
4
4

3i
t1+ p1 t1 p1+
+ r1 1+ 4 l2+ P2 + l2 P2+
2

+ 2
2
+ r 2 0 1 1+ G0 p1+ p1 + 2P2+ P2 + L+
2 p1 + L2 p1

i + +
+
+ T1 p1 P2 + T1 p1 P2 ,
2
225
r1 1+
d
+ b1+ b1 + 2b2+ b2 + f + f iq1 p1+ + iq1+ p1
2
+ 1+ P1 1 P1+ + 22+ P2 22 P2+ ;
(88)
= a+ a +
(89)
meanwhile the explicit expression for the operator B is not essential.

Next, following the procedure of [29,32], we choose the following representation of the
Hilbert space:
(p0 , q1 , p1 , P0 , PG , 1 , P1 , 2 , P2 )|0 = 0,
(90)
and suppose that the vectors and gauge parameters do not depend on G ,

k k
k k k k k k k
| =
(q0 )k1 q1+ 2 p1+ 3 (0 )k4 f + 5 1+ 6 P1+ 7 2+ 8 P2+ 9 b1+ 10 b2+ 11
ki
11
a +1 a +k0 k11k
k (x)|0.
0
(91)
The sum in (91) is taken over k0 , k1 , k2 , k3 , k10 , k11 , running from 0 to infinity, and over k4 ,
k5 , k6 , k7 , k8 , k9 , running from 0 to 1. Then, we derive from the equations that determine the
physical vector, Q | = 0, as well as from the reducible gauge transformations, | = Q |,
a sequence of relations:

Q| = 0,
( + h)| = 0,
(, gh) | = (1, 0),
(92)

| = Q|,
( + h)| = 0,
(, gh) | = (0, 1),
(93)
(1)
(1)
(1)
( + h) = 0,
(, gh)
= (1, 2),
| = Q ,
(94)
(i1)
(i)
(i)
(i)

(95)
= Q ,
( + h) = 0,
(, gh)
= (i, i 1).
The middle equation in (92) presents the equations for the possible values of h,
d 4
(96)
,
2
with n being related to spin, s = n + 1/2. By fixing the value of spin, we also fix the parameter h,
according to (96). Having fixed a value of h, we must substitute it into each of the expressions
(92)(95); see [29] for more details.
As a next step, we have to extract the zero-mode ghosts from the operator Q (88). This operator has the structure

Q = q0 T0 + 0 L 0 + i 1+ q1 1 q1+ p0 + q02 1+ 1 P0 + Q,
(97)
h = n +
226
where
T0 = T0 2q1+ P1 2q1 P1+

i
+ r1+ G0 p1 + T1+ P2 + iL2 p1+ ig0 p1 2 il2 p1+ + t1+ P2
2

+
i
+
+
+ r1 G0 p1+ T1 P2+ iL+
,
p
+
ig
p
+
2
il
p
+
t
P
(98)
0 1
1 2
2 1
2 1
2

+
+ i
+
+
+

L0 = L0 + 2rq1
G0 p1 + T1 P2 + iL2 p1 ig0 p1 2 il2 p1 + t1 P2
2

+
i
+
+
+
+
+
+ 2rq1 G0 p1 T1 P2 iL2 p1 + ig0 p1 + 2 il2 p1 + t1 P2
2

+
+ 1
+
+

+
+ 2r1
G0 P1 + L1 P2 + L2 P1 g0 P1 2 l1 P2 + l2 P1
2

+
1
+
+
+
+
+
2r1 G0 P1 + L1 P2 + L2 P1 g0 P1 2 l2 P1 + l1 P2
2

+

+ 2 i + +
+
+
+ 2
+
2
+ r 1 1 G0 p1 p1 + 2P2 P2 + L2 p1 + L2 p1 + T1 p1 P2 + T1 p1 P2 ,
2
(99)
+
+
+ +
+
Q = q1+ T1 + q1 T1+ + 1+ L1 + 1 L+
+
L
+
i
q
p
+
i
q
p
2 2
2 1 1
1
2 2
2 1 1
+ 2+ 1 P1 + 1+ 2 P1+ 2q1+2 P2 2q12 P2+

3i +
3i
+
+
r1 1+ 2L+
P
+
2L
P
p
+
p
T
T
2
2
1
1
2
2
1
4 1
4

3i +

+ r1 1+ 4 l2+ P2 + l2 P2+
t1 p1 t1 p1+ .
2
(100)
Here, T0 , L 0 , Q are independent of q0 , p0 , 0 , P0 . We also expand the state vector and gauge
parameters in powers of the zero-mode ghosts:
| =

q0k 0k + 0 1k ,

gh mk = (m + k),
(101)

gh (i)km = (i + k + m + 1).
(102)
k=0
(i)

=
q0k (i)k0 + 0 (i)k1 ,
k=0
Following the procedure of [29], we get rid of all the fields except two, |00 , |01 , and, hence,
relations (92)(95), with allowance for (97), (101), (102), yield two independent equations for
these fields:
1

Q00 + T0 , 1+ 1 01 = 0,
2
0

T0 0 + Q01 = 0,
where {A, B} = AB + BA for any quantities A, B.
(103)
(104)
227
Next, due to the fact that the operators Q, T0 , 1+ 1 commute with , we derive from (103),
(104) the equations of motion for the fields with a fixed value of spin:

1
Q00 n + T0 , 1+ 1 01 n = 0,
2

T0 00 n + Q01 n = 0,
(105)
(106)
where the fields |00 , |01 are assumed to obey the relations

00 n = n + (d 4)/2 00 n ,
01 n = n + (d 4)/2 01 n .
(107)
The field equations (105), (106) are Lagrangian ones and can be deduced from the following
action7 :

1 1 +
1
Sn = n 00 Kn T0 00 n +
Kn T0 , 1 1 0 n
2n 0

+ n 00 Kn Q01 n + n 01 Kn Q00 n ,
(108)
where the standard scalar product for the creation and annihilation operators is assumed, and the
operator Kn is the operator K (86), where the following substitution is made: h (n + (d
4)/2).
The equations of motion (105), (106) and the action (108) are invariant with respect to the
gauge transformations

1
00 n = Q00 n + T0 , 1+ 1 10 n ,
2

01 n = T0 00 n + Q10 n ,
(109)
(110)
j
which are reducible, with the gauge parameters |(i) 0 n , j = 0, 1 subject to the same conditions
j
as those for |0 n in (107),

1
(i)00 n = Q(i+1)00 n + T0 , 1+ 1 (i+1)10 n ,
2

(i)10 n = T0 (i+1)00 n + Q(i+1)10 n ,
(0)0
0

0 n = 0 n ,
(0)1
1

0 n = 0 n ,
(111)
(112)
and with a finite number of reducibility stages at imax = n 1 for spin s = n + 1/2.
We now determine the value of the arbitrary parameter m0 , using the condition that the equations (45) for the basic vector | (5) be reproduced. To this end, it is necessary that conditions
(45) be implied by Eqs. (105), (106). Note that the general vector |00 n includes the basic vector
| (5), namely,

0
= |n + |A n , |A n
(113)
0 n
C =P =b+ =b+ =f + =0 = 0.
1
In the next section, we shall demonstrate that due to the gauge transformations and a part of the
equations of motion the vector |A n can be completely removed, so that the resulting equations
of motion have the form
T0 |n = (t0 + m0 )|n = 0,
t1 |n = 0.
7 The action is defined, as usual, up to an overall factor.
(114)
228
The above relations permit one to determine the parameter m0 in a unique way, as follows:

1
1
m0 = m r 2 h = m + r 2 n + (d 4)/2 .
(115)
It should be noted that m0 of the present article is the AdS mass mD of [34] corresponding to
the most-used definition of mass (for more discussions of this point, see [34]).
In the next section, we shall demonstrate that the action actually reproduces the correct equations of motion (12). Thus, we have constructed Lagrangians for fermionic fields of any fixed
spin using the BRST approach.8
6. Reduction to the initial irreducible relations
Let us show that the equations of motion (1), (2) [or equivalently (45)] can be obtained from
the Lagrangian (108) after gauge-fixing and removing the auxiliary fields by using a part of the
equations of motion. Let us start with gauge-fixing.
6.1. Gauge-fixing
Let us consider a field of spin s = n + 1/2. Then we have a reducible gauge theory with n 1
reducibility stages. Due to restriction (107) and the ghost number restriction [see the right-hand
formulae in (102)], the lowest-stage gauge parameters have the form
(n1)0
+ n1 +

(116)
P1 |0 + p1+ |1 0 ,
0 n = p1
(n1)1

(117)
0,
0 n
with the subscripts of the state vectors being associated with the eigenvalues of the corresponding
state vectors (107). It can be verified directly that one can eliminate the dependence on the ghost
P2+ from the gauge function |(n2)00 of the (n 2)th stage (the gauge function |(n2)10 has no
P2+ dependence). It is then possible to verify that one can remove the dependence of |(n3)00 ,
|(n3)10 on P2+ with the help of the remaining gauge parameters (which do not depend on P2+ )
|(n2)00 , |(n2)10 .
We now suppose that we have removed the dependence on P2+ from the gauge functions of the
j
j
(i + 1)th stage |(i+1) 0 , j = 0, 1, i.e., we have 2 |(i+1) 0 = 0. Let us show that these restricted
j
gauge functions can be used to remove the dependence on P2+ from the gauge functions |(i) 0 .
We introduce the following notation for the gauge parameters, related to their decomposition in
ghosts P1+ , P2+ :

(i) j (i) j
+ (i) j
+ (i) j
+ + (i) j

(118)
0 = 00 + P1 01 + P2 02 + P1 P2 03 .
Here and elsewhere, we omit the subscripts of the vectors which are associated with the eigenvalues of the operator (107). Then, using (111) and (112), we find a gauge transformation for
j
j
|(i) 02 and |(i) 03 , being coefficients at P2+ ,

(i)002 = 2q12 (i+1)000 + 2r1+ L2 2l2 (i+1)001 ,
(119)
8 The construction of a Lagrangian describing the propagation of all fermionic fields in the AdS space simultaneously
is analogous to the case of the flat space [29] and we do not consider it here. We only note that the necessary condition for
resolving this problem is to replace in Q , Q, K the parameter h by the operator in an appropriate way and discard
the condition (107) for the fields and gauge parameters.
229

(i)003 = 2q12 (i+1)001 ,
(120)
(i+1)1
(i)1

(i+1)1

(i+1)0
+
2

02 = 2q1
00 + 2r1 L2 2l2
01 + r T1 2t1
01 , (121)

(i)1

03 = 2q12 (i+1)101 .
(122)
Using (119)(122), we can see that the dependence on P2+ in |(i) 0 can be removed with the
help of the gauge transformations. To this end, we should first make gauge transformations with
|(i+1)001 and |(i+1)101 , removing |(i)003 and |(i)103 , respectively. Then we should make
gauge transformation with the parameter |(i+1)000 , removing |(i)002 . Finally, we should make
gauge transformation with the parameter |(i+1)100 , removing |(i)102 . Thus, we have shown
j
that the dependence on P2+ can be eliminated from |(i) 0 .
This procedure works perfectly well until the terms linear in p1+ appear in the gauge functions |(i+1) . When these terms are present, some of the gauge parameters remain unused after
eliminating the P2+ dependence. Due to the presence of r-dependent terms in (119), (121), it is
obvious that if one should make a gauge transformation with such a parameter the terms depending on the ghost P2+ may appear again. Therefore, one should make gauge transformations with
parameters being linear in p1+ or independent of it, before removing the P2+ dependence. The
first gauge function where such a term appear is |(1) . Let us consider gauge transformation
with this gauge function more carefully.
Suppose that the dependence on the ghost P2+ in |(1) has been removed by a gauge transformation. Let us decompose the gauge functions |(1) and |(0) as follows:
(1)0
+ 2
+ +

(123)
0 = ip1 P1 | + p1 (. . .),
(1)1 + 2

(124)
0 = p1 (. . .),
(0)0
+
+
+
+
+
+
+
+

0 = P1 | ip1 |1 i1 p1 P1 |2 iq1 p1 P1 |3
2
i2+ p1+ P1+ |4 + p1+ (. . .) + P2+ (. . .),
(125)
(0)1
+ 2
+
+
+

(126)
0 = ip1 P1 |5 + p1 (. . .) + P2 (. . .).
j
As has been shown, we have to make a gauge transformation with a parameter linear in p1+
(parameters which do not depend on p1+ are absent from |(1) ). Therefore, we use | to make
such a gauge transformation. Since
| = T1+ |,
(127)
we use | to eliminate the dependence on b2+ and f + from |,

b2 | = f | = 0,
(128)
P2+
|(0)
and then we remove the

dependence from
as described above.
Let us turn to the gauge-fixing of the fields. We decompose the field in ghosts P1+ , P2+ by
analogy with the gauge functions,

j j
= + P + j + P + j + P + P + j , j = 0, 1,
(129)
1
2
1 2
0
00
01
02
03
with each |0m being an expansion in p1+ ,
j + k j k
=
p1 0m .
0m
j
k=0
(130)
230
By analogy with the gauge-fixing of |(0) , we have to use |i and then remove the P2+ depenj
dence from the fields |0 .
Substituting (125), (126) into (109), (110), we find a gauge transformation for the fields independent of the ghost p1+ ,
00
+
00
(131)
= L+
1 | + T1 |1 ,
00

+
+
+
+
01 = 1 T1 |2 L1 | |5 + 2 T1 |4 L2 | |3

+ q1+ T1+ |3 T1 | ,
(132)
10
+
= T0 | 2|1 T |5 .
(133)
01
b2+
00 , using | , and then remove the

and f + dependence of |00
Thus, we can first remove the
1
+
00 , using the restricted (128) gauge parameter |. We then remove the b+
b1 dependence of |00
2
10 and | 00 , using the gauge parameters | , | , | , | . After
and f + dependence of |01
5
2
3
4
01
this, we remove the P2+ dependence of the fields. Now, all the gauge parameters have been used,
and we have the following conditions for the fields:
00
00
00
= b2 00
= f 00
= 0,
b1 00
(134)
j 0
j 0
b2 01 = f 01 = 0, j = 0, 1,
(135)
0
1
2 = 2 = 0.
(136)
0
00 )
gh(|00
= 0, it does not depend on ghost coordinates and due to (134) we conclude,

Since
after gauge-fixing, that
00
= |,
(137)
00
where | is the physical field (5).
Let us turn to removing the auxiliary fields with the help of a part of the equations of motion.
6.2. Removing auxiliary fields by equations of motion
Let us decompose Eq. (105) in P1+ , P2+ and consider the coefficients of P2+ -dependent parts:
0
2P2+ P1+ : q12 01
(138)
= 0,

+
+
2 0
0
2P :
(139)
q r L2 2l = 0.
1
00
01
0 contains no terms of an order larger than the first order

Using (138), we conclude that |01
+
in p1 ,
0 00

= + p + 01 .
(140)
01
01
01
1
Substituting (140) to (139), we find

3
0
+ k 0k
=
p1 00 ,
00
(141)
k=0
where
02

= r + L2 2l 00 ,
2
00
01
1
2
03

= r + L2 2l 01 .
2
00
01
1
6
(142)
231
In a similar way, we decompose Eq. (106) and obtain

1
2P2+ P1+ : q12 01
= 0,

1 r
0
+
2 1
T1 2t1 01
= 0.
q1 00 r1+ L2 2l2 01
2P2 :
2
(143)
(144)
These equations have the following solution9 :

1 10

= + p + 11 ,
01
01
01
1
3
1
+ k 1k
=
p1 00 ,
00
(145)
k=1
where
12

= r + L2 2l 10
00
2
01
1
2
13

r
= + L2 2l 11
2
01
00
6 1
00
r
,
T1 2t1 01
4
01
r
.
T1 2t1 01
12
(146)
(147)
00 = |
Let us now consider a part of Eqs. (105), (106) containing the physical field |00
(137),
11
10
+
+ 001

(t0 + m0 )| + iT1+ 00
(148)
+ L+
010 = 0,
1 01 r L2 2l2

10
+
+ 011
+ 001

q1 : t1 | + iT1 000 + L1 010 01 = 0,
(149)

11
+
000
+
+ 010
+ 000
+
1 : l1 | iT1 001 + L1 011 i 00 2r L2 2l2 012
10
001
3r +
(150)
+
T0 01
= 0,
T1 2t1+ 010
4

010
011
000 000
+ i 000
+ L+
2+ : l2 | iT1+ 002
(151)
1 012 011 = 0,
00 and | 01 as follows:
where we have decomposed the fields |01
00
001
000
000

01
00

+
+
+

= q + 011 + + 010 + + 010 .
= q
01
010 + 1 011 + 2 012 ,
00
000
001
002
1
1
1
2
(152)
Acting on (149) by the operators f and b2 and taking into account the gauge-fixing conditions
(134), (135), we obtain, respectively,
011
011
f T1+ 000
(153)
= 0,
b2 T1+ 000
= 0.
These equations yield

011

000 = 0.
Acting on (151) by f and b2 , we obtain
010
010
f T1+ 002
= b2 T1+ 002
= 0
We now act on Eq. (148) by the operator f ,
11
f T1+ 00
= 0,
(154)
010

002 = 0.
(155)
(156)
9 The term | 10 corresponding to (p + )0 is absent in the sum of (145) due to the ghost number restriction (101).
1
00
232
11 as a power series in f + ,
and then presenting the vector |00

11 11
= a + f + b11 ,
00
00
00
(157)
we have

11
a = t + b11 .
00
00
1
(158)
Substituting this result into Eq. (148) and acting twice by the operator b2 on the resulting equations, we arrive at
11
b2 b00
(159)
= 0.
Finally, Eq. (148) acquires the form
11

+
+ 10
+ 001

010 = 0.
(t0 + m0 )| + 2iL+
2 b00 + L1 01 r L2 2l2
(160)
11 , | 10 , | 001 as a power series in b+ ,

Let us present the state vectors |b00
01
010
1
n2
11
+ k 11 k
b =
b1 b00 ,
00
k=0
n1
10
+ k 10 k
=
b
,
01
01
k=0
n2
001
+ k 001 k

=
b
, (161)
010
010
k=0
and then, presenting equations (160) and (149) as power series in b1+ and b2+ , we have
10 n1
n
b1+ : m1 01
= 0,
10 n2
10 n1
+ n1
b1
: m1 01
= l1+ 01
,
10 k1
10 k
11 k
001 k
+ k
1 k n 2,
b1 : m1 01
= l1+ 01
2il2+ b00
+ rl2+ 010
,

+ 0
+ 11 0
+ 10 0
+ 001 0
b1 : (t0 + m0 )| = 2il2 b00 l1 01 + rl2 010 ,
k 11 k ir 001 k
0 k n 2, b2+ b1+ : b00
= 010 ,
2
001 n2 10 n1
+ n1

: m1 010
= 01
,
b1
001 k1 10 k
001 k
+ k
1 k n 2,
b1 : m1 010
= 01 l1+ 010
,
10 0

+ 0

001 0
b1 : t1 | = 01 l1+ 010
.

(162)
(163)
(164)
(165)
(166)
(167)
(168)
(169)
10 n1 = | 10 n2 = 0. Substituting this result into (167) and

Using (162) and (163), we have |01
01
001 n2 = | 001 n3 = 0. Turning to Eq. (166) for
then into (168) for k = n 2, we obtain |010
010
11 n2 = |b11 n3 = 0. We now repeat the procedure
k = n 2 and k = n 3, we conclude that |b00
00
starting from Eq. (164) for k = n 2 and k = n 3. Finally, we obtain
(t0 + m0 )| = 0,
10 001
=
= 0,
01
010
t1 | = 0,
11
b = 0
00
11
= 0.
00
(170)
(171)
Using (170), we can see that the physical state (5) satisfies (1), (2), or equivalently (45), provided
that condition (115) is taken into account.
Let us now turn to Eqs. (150), (151). Taking into account (154), (155), (171), we can see that
these two equations read
010
000
+
+ 000

+ L+
012 = 0,
l1 | iT1+ 001
1 011 2r L2 2l2
000

+ 000

= 0.
l2 | + L
012
011
233
(172)
(173)
011 = 0 in (154) is taken into account. ThereThey are analogous to (148) and (149), where |000
fore, we can repeat the procedure that has been carried out with Eqs. (148), (149), and thus we
arrive at the conclusion that
l1 | = l2 | = 0,
010 000 000

=
=
= 0.
001
012
011
(174)
(175)
Equations (174) are consequences of (170); they do not impose any additional restrictions on the
physical state (5). Collecting (154), (155), (171), (175) and (152), we have
10 11 00 01
= = = = 0.
(176)
00
01
01
00
Observing (140), (141), (142) and (145), (146), (147), we can see that it remains to show that
01 = | 11 = 0. To prove this fact, we decompose (105) and (106) in the ghosts p + , P + and
|01
1
1
01
consider the equations which are the coefficients at (p1+ )0 P1+ . With allowance for (176), these
equations read
01
01
11
11
iT1+ 01
(177)
2+ p1 01
= 0,
iT1+ 01
2+ p1 01
= 0,
01 , | 11 in ghost coordinates,
respectively. Decomposing |01
01
01

11

= q + 011 + + 010 + + 010 ,
= q + 111 + + 110 + + 110 ,
01
010
01
010
011
012
011
012
1
1
2
1
1
2
(178)
and substituting the result into (177), we obtain equations for the coefficients of (178) in the form
j 1i
j 1i
T1+ 01n = 0 01n = 0.
(179)
01 11
Therefore, all the coefficient in decompositions (178) of 01 , 01 are equal to zero, and we
conclude that
01 11
= = 0.
(180)
01
01
Thus, we have shown that after the gauge fixing (134)(136) all the auxiliary fields become equal
00 (137) obeys Eqs. (1), (2).
to zero, (176), (180), and the physical state | = |00
Let us now consider some examples of the Lagrangian construction procedure.
7. Examples
Here, we shall illustrate the general procedure of gauge-invariant Lagrangian construction by
using the examples of fermionic fields of spin 1/2 and 3/2.
7.1. Spin-1/2 field
For a fermionic field of spin s = 12 , we have h = 2 d2 . Then the only non-vanishing vector
|00 0 subject to condition (107) and having the proper ghost number (92) has the form
0
0
= (x)|0,
= 0 0| + (x) 0 .
(181)
0 0
0 0
234
Then, due to (75), (86) for K0 = |00| + , (115), the action implied by (108) has the form

0
0
1
d
d

2
2 .
S0 = 0 0 K0 T0 0 0 = d x |g| i m r
(182)
2
Here, we have applied the definition (14) for the conventional gamma-matrices and have introduced the Dirac-conjugate spinor , = + 0 . Thus, we can see that the action (182)
reproduces Eq. (1) for n = 0, which corresponds to spin-1/2 field.
7.2. Spin-3/2 field
1
In the case of a spin-3/2 field, we have n = 1, h = 1 d/2, m0 = m + r 2 (d 2)/2.

Since imax = 0, the corresponding Lagrangian formulation is an irreducible gauge theory. Due
to gh(|10 1 ) = 2, the non-vanishing fields |00 1 , |01 1 and the gauge parameter |00 1 , for
|10 1 0, possess the following Grassmann grading and ghost number distributions:

(, gh) 00 1 = (1, 0),
(, gh) 01 1 = (1, 1),
(, gh) 00 1 = (0, 1).
(183)
These conditions determine the dependence of the fields and gauge parameter on the oscillator
variables in a unique form:
0

= ia + (x) + f + (x) + b+ (x) |0,
0 1
1
1

= P + (x) + ip + 1 (x) |0,
(184)
0 1
1
1
0

+
= 0| i (x)a + + (x) f + + (x)b1 0 ,
1 0
1

(185)
= 0| + (x) P1 1+ ip1 0 ,
1 0
0

= P + 1 (x) ip + 2 (x) |0.
0 1
1
1
(186)
Substituting (184), (185) into (108), we find the action (up to an overall factor) for a spin-3/2
field interacting with the AdS background:

S1 = d d x |g| i m0 + i 1

m0
r(d 1)
+ (d 2)
i + m0
1
m1
2m1
2

M
m0
i m0 2m1 m1
2
m1
m1

M2
+ i + m0 + 1 + + m0 +
m1

m0
1 i + (d 2)
(187)
,
m1
where M 2 = m20 14 r(d 1)(d 2). To obtain the action (187), we have used the expressions
for the operators K1 (75), (86). Substituting (184)(186) into (109), (110), we find the gauge
transformations
235
= 1 + i 2 ,
= 2 ,
= m1 1 ,
(188)

r
= i m0 1 22 ,
(189)
1 = i + m0 2 (d 1)1 .
2
Let us present the action in terms of one physical field . To this end, we get rid of the fields
, , by using their gauge transformations and the gauge parameters 1 , 2 , respectively. Having
expressed the field , using the equation of motion = i , we can see that the terms with
the Lagrangian multiplier 1 turn to zero. As a result, we obtain

LRS = i m0 i ( + ) + i + m0 .
(190)
This is a generalization of the RaritaSchwinger Lagrangian to a d-dimensional AdS space.
8. Conclusion
We have constructed a gauge-invariant Lagrangian formulation of half-integer totally symmetric higher spin fields in the AdS space of any dimension in the metric-like formulation.
The results of this study are most general and apply to both massive and massless fermionic
higher spin fields in the AdS, Minkowski, and dS10 spaces.
Starting from embedding the fermionic higher spin fields into vectors of an auxiliary Fock
space, we treat the fields as components of these Fock-space vectors, and, as a result, we reformulate the theory in terms of such vectors. We realize the conditions that define an irreducible
representation of the AdS group with a given mass and spin in terms of differential operators
acting in this Fock space. The mentioned conditions are interpreted as constraints imposed on
the Fock space vectors and generate a closed higher spin nonlinear symmetry superalgebra being
the basic object of this study.
It is shown that the derivation of a correct Lagrangian formulation requires a transition to another constraint basis for the original symmetry algebra, which is algebraically equivalent to the
initial basis and does not contain the constraints that define an irreducible representation of the
AdS group. This set of constraints provides an additive extension of the initial algebra to another
operator algebra. Then, as shown in [33], it is necessary to deform the initial algebra. As a result,
one obtains a nonlinear superalgebra of enlarged constraints, whose construction by means of the
additional operators realizes a special conversion of the initial system of first- and second-class
operator constraints into a system of first-class ones with a preservation of the initial algebraic
structure for the deformed constraints (see [37] for the elaboration of conversion methods). Due
to the nonlinearity of the underlying algebra of the enlarged constraints, the corresponding BRST
operator is defined ambiguously, and we construct it in an exact form with a non-vanishing term
of third order in powers of ghosts for the case of a supersymmetric ordering of the constraints
in commutator relations. It is shown that the resulting BRST operator yields a consistent Lagrangian dynamics for fermionic fields of any spin. The corresponding Lagrangian formulation
is constructed in a concise form in terms of a Fock space and proves to be a reducible gauge
theory with a finite number of reducibility stages, the number growing with the spin value. It
is interesting to observe that the methods developed for the quantization of gauge theories turn
out to be extremely efficient for deriving classical gauge-invariant Lagrangians for higher spin
field theories, thus reflecting one more side of the BVBFV duality concept [38], which permits
1
10 In the case of HS fields on dS space the corresponding Lagrangians lose the property of reality due to r 2 .
236
one to construct, by means of a Hamiltonian BFVBRST charge, the objects used in Lagrangian
formalism.
We have proved that the Lagrangian equations of motion (105), (106), after a partial gaugefixing, reproduce the equations corresponding to the relations determining the irreducible representation of the AdS group. This proof completes the derivation of a correct gauge-invariant
Lagrangian formulation for higher spin fermionic fields in the AdS space. As examples demonstrating the general scheme of Lagrangian construction, we have obtained gauge-invariant Lagrangian descriptions of fields with spin 1/2 and spin 3/2 in an explicit form. In principle, the
constructed description permits one to obtain explicit Lagrangians for any other half-integer-spin
fields.
The basic results of the present work are given by relations (108), where the Lagrangian
action for a field with an arbitrary half-integer spin is constructed, and by (109)(112), where
the gauge transformations for the fields are presented as well as the sequence of reducible gauge
transformations for the gauge parameters.
Concluding, we shall discuss two points. First, the gauge-invariant formulation for massive
higher spin field theories in the AdS space fields can have very interesting applications in the
calculation of a quantum effective action for higher spin massive fields in the AdS space. The
Lagrangians of all such models have a gauge-invariant kinetic term and a mass term violating the
gauge invariance. Such a structure of a Lagrangian leads to some problems of quantum calculations in a curved spacetime. To avoid these problems, it is natural to construct a gauge-invariant
formulation by using appropriate Stuckelberg fields. Then, one can impose gauge-fixing conditions removing the gauge degeneration of the kinetic term in the Lagrangian of the basic fields
and apply the standard techniques for the calculation of an effective action (see the examples
of such calculations in [39]). We emphasize that the BRST approach leads automatically to a
gauge-invariant Lagrangian with the entire set of appropriate Stckelberg fields. As a result, one
has a basis for constructing the effective action of massive higher spin fields in the AdS space.
Second, the BRST approach can provide a systematic method of constructing interaction vertices
for higher spin fields in the AdS space [30]. Therefore, one can hope that the BRST construction
developed in this paper will be useful in deriving interaction vertices for massive higher spin
fermionic fields in the AdS space.
Acknowledgements
A.A.R. is grateful to M. Grigoriev and P. Moshin for useful discussions. The work of I.L.B.
and V.A.K. was partially supported by the INTAS grant, project INTAS-05-7928, the RFBR
grant, project No. 06-02-16346, and the grant for LRSS, project No. 4489.2006.2. The work of
I.L.B. was supported in part by the DFG grant, project No. 436 RUS 113/669/0-3, and the joint
RFBR-DFG grant, project No. 06-02-04012. The work of V.A.K. was partially supported by the
joint DAADMikhail Lomonosov Programme (Referat 325, Kennziffer A/06/16774).
References
[1] M. Vasiliev, Higher spin gauge theories in various dimensions, Fortschr. Phys. 52 (2004) 702717, hep-th/0401177;
D. Sorokin, Introduction to the classical theory of higher spins, AIP Conf. Proc. 767 (2005) 172202, hep-th/
0405069;
N. Bouatta, G. Compre, A. Sagnotti, An introduction to free higher-spin fields, hep-th/0409068;
A. Sagnotti, E. Sezgin, P. Sundell, On higher spin with a strong Sp(2) conditions, hep-th/0501156;
X. Bekaert, S. Cnockaert, C. Iazeolla, M.A. Vasiliev, Nonlinear higher spin theories in various dimensions, hepth/0503128.
237
[2] I.L. Buchbinder, V.A. Krykhtin, V.D. Pershin, On consistent equations for massive spin-2 field coupled to gravity
in string theory, Phys. Lett. B 466 (1999) 216226, hep-th/9908028;
I.L. Buchbinder, D.M. Gitman, V.A. Krykhtin, V.D. Pershin, Equations of motion for massive spin 2 field coupled
to gravity, Nucl. Phys. B 584 (2000) 615640, hep-th/9910188;
I.L. Buchbinder, D.M. Gitman, V.D. Pershin, Causality of massive spin 2 field in external gravity, Phys. Lett. B 492
(2000) 161170, hep-th/0006144.
[3] S. Deser, A. Waldron, Gauge invariances and phases of massive higher spins in (A)dS, Phys. Rev. Lett. 87 (2001)
031601, hep-th/0102166;
S. Deser, A. Waldron, Partial masslessness of higher spins in (A)dS, Nucl. Phys. B 607 (2001) 577604, hep-th/
0103198;
S. Deser, A. Waldron, Null propagation of partially massless higher spins in (A)dS and cosmological constant
speculations, Phys. Lett. B 513 (2001) 137141, hep-th/0105181;
K. Hallowell, A. Waldron, Constant curvature algebras and higher spin action generating functions, Nucl. Phys.
B 724 (2005) 453486, hep-th/0505255;
E.D. Skvortsov, M.A. Vasiliev, Geometric formulation for partially massless fields, Nucl. Phys. B 756 (2006) 117
147, hep-th/0601095.
[4] R.R. Metsaev, Massive totally symmetric fields in AdS(d), Phys. Lett. B 590 (2004) 95104, hep-th/0312297;
R.R. Metsaev, Mixed-symmetry massive fields in AdS(5), Class. Quantum Grav. 22 (2005) 27772796, hep-th/
0412311;
R.R. Metsaev, Cubic interaction vertices of massive and massless higher spin fields, Nucl. Phys. B 759 (2006)
147201, hep-th/0512342.
[5] S.M. Klishevich, Massive fields with arbitrary integer spin in symmetrical Einstein space, Class. Quantum Grav. 16
(1999) 29152927, hep-th/9812005;
S.M. Klishevich, On electromagnetic interaction of massive spin-2 particle, Phys. At. Nucl. 61 (1998) 15271537,
hep-th/9708150;
S.M. Klishevich, Massive fields with arbitrary half-integer spin in constant electromagnetic field, Int. J. Mod. Phys.
A 15 (2000) 609624, hep-th/9811030;
S.M. Klishevich, Massive fields with arbitrary integer spin in homogeneous electromagnetic field, Int. J. Mod. Phys.
A 15 (2000) 535, hep-th/9810228;
S.M. Klishevich, Interaction of massive integer-spin fields, hep-th/0002024.
[6] Yu.M. Zinoviev, On massive high spin particles in (A)dS, hep-th/0108192;
Yu.M. Zinoviev, On massive mixed symmetry tensor fields in Minkowski space and (A)dS, hep-th/0211233;
Yu.M. Zinoviev, Massive spin-2 supermultiplets, hep-th/0206209;
Yu.M. Zinoviev, Massive supermultiplets with spin 3/2, hep-th/0703118;
Yu.M. Zinoviev, Massive N = 1 supermultiplets with arbitrary superspins, hep-th/07041535.
[7] N. Beisert, M. Bianchi, J.F. Morales, H. Samtleben, Higher spin symmetries and N = 4 SYM, JHEP 0407 (2004)
058, hep-th/0405057;
A.C. Petkou, Holography, duality and higher spin fields, hep-th/0410116;
M. Bianchi, P.J. Heslop, F. Riccioni, More on La Grande Bouffe: Towards higher spin symmetry breaking in AdS,
JHEP 0508 (2005) 088, hep-th/0504156;
P.J. Heslop, F. Riccioni, On the fermionic Grande Bouffe: More on higher spin symmetry breaking in AdS/CFT,
JHEP 0510 (2005) 060, hep-th/0508086;
M. Bianchi, V. Didenko, Massive higher spin multiplets and holography, hep-th/0502220.
[8] I.L. Buchbinder, S.J. Gates, W.D. Linch, J. Phillips, New 4d, N = 1 superfiled theory: Model of free massive
superspin-3/2 multiplet, Phys. Lett. B 535 (2002) 280288, hep-th/0108200;
I.L. Buchbinder, S.J. Gates, W.D. Linch, J. Phillips, Dynamical superfiled theory of free massive superspin-1 multiplet, Phys. Lett. B 549 (2002) 229236, hep-th/0207243;
I.L. Buchbinder, S.J. Gates, S.M. Kuzenko, J. Phillips, Massive 4D, N = 1 superspin 1 and 3/2 multiplets and their
dualities, JHEP 0502 (2005) 056, hep-th/0501199;
S. Fedoruk, J. Lukiersky, Massive relativistic models with bosonic counterpart of supersymmetry, Phys. Lett. B 632
(2006) 371378, hep-th/0506086.
[9] P. de Medeiros, Massive gauge-invariant field theories on space of constant curvature, Class. Quantum Grav. 21
(2004) 25712593, hep-th/0311254.
[10] L. Brink, R.R. Metsaev, M.A. Vasiliev, How massless are massless fields in AdSd , Nucl. Phys. B 586 (2000) 183
205, hep-th/0005136;
238
[11]
[12]
[13]
[14]
[15]
[16]
[17]

K.B. Alkalaev, M.A. Vasiliev, N = 1 supersymmetric theory of higher spin gauge fields in AdS(5) at the cubic level,
Nucl. Phys. B 655 (2003) 5792, hep-th/0206068;
K.B. Alkalaev, Two-column higher spin massless fields in AdS(d), Theor. Math. Phys. 140 (2004) 12531263,
hep-th/0311212;
O.V. Shaynkman, I.Yu. Tipunin, M.A. Vasiliev, Unfolded form of conformal equations in M dimensions and o(M +
2)-modules, Rev. Math. Phys. 18 (2006) 823886, hep-th/0401086;
K.B. Alkalaev, O.V. Shaynkman, M.A. Vasiliev, On the frame-like formulation of mixed symmetry massless fields
in (A)dS(d), Nucl. Phys. B 692 (2004) 363393, hep-th/0311164;
K.B. Alkalaev, O.V. Shaynkman, M.A. Vasiliev, Lagrangian formulation for free mixed-symmetry bosonic gauge
fields in (A)dS(d), JHEP 0508 (2005) 069, hep-th/0501108;
K.B. Alkalaev, O.V. Shaynkman, M.A. Vasiliev, Frame-like formulation for free mixed-symmetry bosonic massless
higher spin fields in AdS(d), hep-th/0601225.
R.R. Metsaev, Free totally (anti)symmetric massless fermionic fields in d-dimensional anti-de Sitter space, Class.
Quantum Grav. 14 (1997) L115L121, hep-th/9707066;
R.R. Metsaev, Fermionic fields in the d-dimensional anti-de Sitter spacetime, Phys. Lett. B 419 (1998) 4956,
hep-th/9802097;
R.R. Metsaev, Arbitrary spin massless bosonic fields in d-dimensional anti-de Sitter space, hep-th/9810231;
R.R. Metsaev, Light-cone form of field dynamics in anti-de Sitter spacetime and AdS/CFT correspondence, Nucl.
Phys. B 563 (1999) 295348, hep-th/9906217;
R.R. Metsaev, Massless arbitrary spin fields in AdS(5), Phys. Lett. B 531 (2002) 152160, hep-th/0201226.
R.R. Metsaev, Massless mixed symmetry bosonic free fields in d-dimensional anti-de Sitter spacetime, Phys. Lett.
B 354 (1995) 7884;
R.R. Metsaev, Fermionic fields in the d-dimensional anti-de Sitter spacetime, Phys. Lett. B 419 (1998) 4956,
hep-th/9802097.
D. Francia, A. Sagnotti, Free geometric equations for higher spins, Phys. Lett. B 543 (2002) 303310, hep-th/
0207002;
D. Francia, A. Sagnotti, On the geometry of higher spin gauge fields, Class. Quantum Grav. 20 (2003) S473S486,
hep-th/0212185;
D. Francia, A. Sagnotti, Minimal local Lagrangians for higher-spin geometry, Phys. Lett. B 624 (2005) 93104,
hep-th/0507144;
A. Sagnotti, M. Tsulaia, On higher spins and the tensionless limit of string theory, Nucl. Phys. B 682 (2004) 83116,
hep-th/0311257;
A. Fotopoulos, K.L. Panigrahi, M. Tsulaia, Lagrangian formulation of higher spin theories on AdS, Phys. Rev. D 74
(2006) 085029, hep-th/0607248;
D. Francia, J. Mourad, A. Sagnotti, Current exchanges and unconstrained higher spins, hep-th/0701163;
I.L. Buchbinder, A.V. Galajinsky, V.A. Krykhtin, Quartet unconstrained formulation for massless higher spin fields,
hep-th/0702161.
F. Kristiansson, P. Rajan, Scalar field corrections to AdS4 gravity from higher spin gauge theory, JHEP 0304 (2003)
009, hep-th/0303202.
X. Bekaert, N. Boulaunger, S. Cnockraert, Spin three gauge filed theory revised, JHEP 0601 (2006) 052, hep-th/
0508048.
G. Bonelli, On the tensionless limit of bosonic strings, infinite symmetries and higher spins, Nucl. Phys. B 669
(2003) 159172, hep-th/0305155;
G. Barnich, G. Bonelli, M. Grigoriev, From BRST to light-cone description of higher spin gauge fields, hep-th/
0502232.
M. Plyushchay, D. Sorokin, M. Tsulaia, Higher spins from tensorial charges and OSp(N |2n) symmetry, JHEP 0304
(2003) 013, hep-th/0301067;
M. Plyushchay, D. Sorokin, M. Tsulaia, GL flatness of OSp(1|2n) and higher spin field theory from dynamics in
tensorial space, hep-th/0310297;
I. Bandos, X. Bekaert, J.A. Azcarraga, D. Sorokin, M. Tsulaia, Dynamics of higher spin fields and tensorial space,
JHEP 0505 (2005) 031, hep-th/0501113;
E. Ivanov, J. Lukiersky, Higher spins from nonlinear realizations of OSp(1|8), Phys. Lett. B 624 (2005) 304315,
hep-th/0505216;
S. Fedoruk, E. Ivanov, Master higher spin particle, Class. Quantum Grav. 23 (2006) 51955214, hep-th/0604111;
S. Fedoruk, E. Ivanov, J. Lukiersky, Massless higher spin D = 4 superparticle with both N = 1 supersymmetry and
its bosonic counterpart, Phys. Lett. B 641 (2006) 226236, hep-th/0606053.
239
[18] G. Barnich, M. Grigoriev, A. Semikhatov, I. Tipunin, Parent field theory and unfolding in BRST first-quantized
terms, Commun. Math. Phys. 260 (2005) 147181, hep-th/0406192;
G. Barnich, M. Grigoriev, Parent form for higher spin fields on anti-de Sitter space, hep-th/0602166;
M. Grigoriev, Off-shell gauge fields from BRST quantization, hep-th/0605089.
[19] C. Fronsdal, Singletons and massless, integer-spin fields on de Sitter space, Phys. Rev. D 20 (1979) 848856;
J. Fang, C. Fronsdal, Massless, half-integer-spin fields in de Sitter space, Phys. Rev. D 22 (1980) 13611367;
M.A. Vasiliev, Free massless fermionic fields of arbitrary spin in D-dimensional anti-de Sitter space, Nucl. Phys.
B 301 (1988) 2651;
V.E. Lopatin, M.A. Vasiliev, Free massless bosonic fields of arbitrary spin in D-dimensional de Sitter space, Mod.
Phys. Lett. A 3 (1998) 257265.
[20] E.S. Fradkin, M.A. Vasiliev, Cubic interaction in extended theories of massless higher spin fields, Nucl. Phys. B 291
(1987) 141;
E.S. Fradkin, M.A. Vasiliev, Candidate to the role of higher spin symmetry, Ann. Phys. 177 (1987) 63;
E.S. Fradkin, M.A. Vasiliev, On the gravitational interaction of massless higher spin fields, Phys. Lett. B 189 (1987)
8995;
M.A. Vasiliev, Consistent equation for interacting gauge fields of all spins in (3 + 1)-dimensions, Phys. Lett. B 243
(1990) 378382;
M.A. Vasiliev, Cubic interactions of bosonic higher spin gauge fields in AdS(5), Nucl. Phys. B 616 (2001) 106162,
hep-th/0106200;
M.A. Vasiliev, Cubic interactions of bosonic higher spin gauge fields in AdS(5), Nucl. Phys. B 652 (2003) 407,
Erratum;
M.A. Vasiliev, Algebraic aspects of the higher spin problem, Phys. Lett. B 257 (1991) 111118;
M.A. Vasiliev, More on equations of motion for interacting massless fields of all spins in (3 + 1)-dimensions, Phys.
Lett. B 285 (1992) 225234;
M.A. Vasiliev, Class. Quantum Grav. 8 (1991) 1387;
M.A. Vasiliev, Nonlinear equations for symmetric massless higher spin fields in (A)dS(d), Phys. Lett. B 567 (2003)
139151, hep-th/0304049.
[21] E. Sezgin, P. Sundell, Doubletons and 5-D higher spin gauge theory, JHEP 0109 (2001) 036, hep-th/0105001;
E. Sezgin, P. Sundell, Towards massless higher spin extension of D = 5, N = 8 gauged supergravity, JHEP 0109
(2001) 025, hep-th/0107186;
E. Sezgin, P. Sundell, Analysis of higher spin field equations in four-dimensions, JHEP 0207 (2002) 055, hep-th/
0205132;
E. Sezgin, P. Sundell, Holography in 4D (super) higher spin theories and a test via cubic scalar couplings, JHEP 0507
(2005) 044, hep-th/0305040;
E. Sezgin, P. Sundell, An exact solution of 4d higher-spin gauge theory, Nucl. Phys. B 762 (2007) 137, hep-th/
0508158;
J. Engquist, E. Sezgin, P. Sundell, On N = 1, N = 2, N = 4 higher spin gauge theories in four-dimensions, Class.
Quantum Grav. 19 (2002) 61756196, hep-th/0207101.
[22] J.M. Maldacena, The large N limit of superconformal field theories and supergravity, Adv. Theor. Math. Phys. 2
(1998) 231, hep-th/9711200.
[23] M. Fierz, W. Pauli, On relativistic wave equations for particles of arbitrary spin in an electromagnetic field, Proc. R.
Soc. London, Ser. A 173 (1939) 211232;
L.P.S. Singh, C.R. Hagen, Lagrangian formulation for arbitrary spin. 1. The bosonic case, Phys. Rev. D 9 (1974)
898909;
L.P.S. Singh, C.R. Hagen, Lagrangian formulation for arbitrary spin. 2. The fermionic case, Phys. Rev. D 9 (1974)
910920.
[24] E.S. Fradkin, G.A. Vilkovisky, Quantization of relativistic systems with constraints, Phys. Lett. B 55 (1975) 224
226;
I.A. Batalin, G.A. Vilkovisky, Relativistic S-matrix of dynamical systems with boson and fermion constraints, Phys.
Lett. B 69 (1977) 309312;
I.A. Batalin, E.S. Fradkin, Operator quantization of relativistic dynamical systems subject to first class constraints,
Phys. Lett. B 128 (1983) 303.
[25] I.A. Batalin, E.S. Fradkin, Operator quantization method and abelization of dynamical systems subject to first class
constraints, Riv. Nuovo Cimento 9 (10) (1986) 1;
I.A. Batalin, E.S. Fradkin, Operator quantization of dynamical systems subject to constraints. A further study of the
construction, Ann. Inst. H. Poincar A 49 (1988) 145;
240
M. Henneaux, C. Teitelboim, Quantization of Gauge Systems, Princeton Univ. Press, 1992.

[26] S. Ouvry, J. Stern, Gauge fields of any spin and symmetry, Phys. Lett. B 177 (1986) 335340;
A.K.H. Bengtsson, A unified action for higher spin gauge bosons from covariant string theory, Phys. Lett. B 182
(1986) 321325;
W. Siegel, B. Zwiebach, Gauge string fields from light cone, Nucl. Phys. B 282 (1987) 125;
W. Siegel, Gauging Ramond string fields via OSp(1, 1|2), Nucl. Phys. B 284 (1987) 632.
[27] A. Pashnev, M. Tsulaia, Description of the higher massless irreducible integer spins in the BRST approach, Mod.
Phys. Lett. A 13 (1998) 18531864, hep-th/9803207;
C. Burdik, A. Pashnev, M. Tsulaia, On the mixed symmetry irreducible representations of the Poincare group in the
BRST approach, Mod. Phys. Lett. A 16 (2001) 731746, hep-th/0101201.
[28] I.L. Buchbinder, A. Pashnev, M. Tsulaia, Lagrangian formulation of the massless higher integer spin fields in the
AdS background, Phys. Lett. B 523 (2001) 338346, hep-th/0109067;
X. Bekaert, I.L. Buchbinder, A. Pashnev, M. Tsulaia, On higher spin theory: Strings, BRST, dimensional reductions,
Class. Quantum Grav. 21 (2004) 14571464, hep-th/0312252.
[29] I.L. Buchbinder, V.A. Krykhtin, A. Pashnev, BRST approach to Lagrangian construction for fermionic massless
higher spin fields, Nucl. Phys. B 711 (2005) 367391, hep-th/0410215.
[30] I.L. Buchbinder, A. Fotopoulos, A.C. Petkou, M. Tsulaia, Constructing the cubic interaction vertex of higher spin
gauge fields, Phys. Rev. D 74 (2006) 105018, hep-th/0609082.
[31] I.L. Buchbinder, V.A. Krykhtin, Gauge invariant Lagrangian construction for massive bosonic higher spin fields in
D dimensions, Nucl. Phys. B 727 (2005) 536563, hep-th/0505092.
[32] I.L. Buchbinder, V.A. Krykhtin, L.L. Ryskina, H. Takata, Gauge invariant Lagrangian construction for massive
higher spin fermionic fields, Phys. Lett. B 641 (2006) 386392, hep-th/0603212.
[33] I.L. Buchbinder, V.A. Krykhtin, P.M. Lavrov, Gauge invariant Lagrangian formulation of higher massive bosonic
field theory in AdS space, Nucl. Phys. B 762 (2007) 344376, hep-th/0608005.
[34] R.R. Metsaev, Gauge invariant formulation of massive totally symmetric fermionic fields in (A)dS space, Phys.
Lett. B 643 (2006) 205212, hep-th/0609029.
[35] K. Schoutens, A. Sevrin, P. van Nieuwenhuizen, Quantum BRST charge for quadratically nonlinear Lie algebras,
Commun. Math. Phys. 124 (1989) 87103;
I.L. Buchbinder, P.M. Lavrov, Classical BRST charge for non-linear algebras, hep-th/0701243.
[36] I.L. Buchbinder, A. Pashnev, M. Tsulaia, Massless higher spin fields in the AdS background and BRST construction
for nonlinear algebras, hep-th/0206026;
C. Burdik, O. Navratil, A. Pashnev, On the Fock space realizations of nonlinear algebras describing the high spin
fields in AdS spaces, hep-th/0206027.
[37] L.D. Faddeev, S.L. Shatoshvili, Realization of the Schwinger term in the Gauss low and the possibility of correct
quantization of a theory with anomalies, Phys. Lett. B 167 (1986) 225338;
I.A. Batalin, E.S. Fradkin, T.E. Fradkina, Another version for operatorial quantization of dynamical systems with
irreducible constraints, Nucl. Phys. B 314 (1989) 158174;
I.A. Batalin, I.V. Tyutin, Existence theorem for the effective gauge algebra in the generalized canonical formalism
and Abelian conversion of second class constraints, Int. J. Mod. Phys. A 6 (1991) 32553282;
E. Egorian, R. Manvelyan, Quantization of dynamical systems with first and second class constraints, Theor. Math.
Phys. 94 (1993) 241252.
[38] M. Grigoriev, P.H. Damgaard, Superfield BRST charge and the master action, Phys. Lett. B 474 (2000) 323330,
hep-th/9911092;
G. Barnich, M. Grigoriev, Hamiltonian BRST and BatalinVilkovisky formalisms for second quantization of gauge
theories, Commun. Math. Phys. 254 (2005) 581601, hep-th/0310083;
D.M. Gitman, P.Yu. Moshin, A.A. Reshetnyak, Local superfield Lagrangian BRST quantization, J. Math. Phys. 46
(2005) 072302-01072302-24, hep-th/0507160;
D.M. Gitman, P.Yu. Moshin, A.A. Reshetnyak, An embedding of the BV quantization into an N = 1 local superfield
formalism, Phys. Lett. B 621 (2005) 295308, hep-th/0507049.
[39] F.A. Dilkes, M.J. Duff, J.K. Liu, H. Sati, Quantum discontinuity between zero and infinitesimal graviton mass with
a Lambda term, Phys. Rev. Lett. 87 (2001) 041301, hep-th/0102093;
M.J. Duff, J.K. Liu, H. Sati, Quantum M 2 2/3 discontinuity for massive gravity with a Lambda term, Phys.
Lett. B 515 (2001) 156160, hep-th/0105008;
I.L. Buchbinder, G. de Berredo-Pexoto, I.L. Shapiro, Quantum effects in softly broken gauge theories in curved
spacetimes, Phys. Lett. B 649 (2007) 454462, hep-th/0703189.
Electronic properties of graphene

with a topological defect
Yu.A. Sitenko a,b, , N.D. Vlasii a,b
a Bogolyubov Institute for Theoretical Physics, National Academy of Sciences, 03680 Kyiv, Ukraine
b Physics Department, National Taras Shevchenko University of Kyiv, 03127 Kyiv, Ukraine
Received 15 May 2007; accepted 6 June 2007

Abstract
Various types of topological defects in graphene are considered in the framework of the continuum model
for long-wavelength electronic excitations, which is based on the DiracWeyl equation. The condition for
the electronic wave function is specified, and we show that a topological defect can be presented as a
pseudomagnetic vortex at the apex of a graphitic nanocone; the flux of the vortex is related to the deficit
angle of the cone. The cases of all possible types of pentagonal defects, as well as several types of heptagonal defects (with the numbers of heptagons up to three, and six) are analyzed. The density of states and the
ground state charge are determined.
PACS: 11.10.-z; 73.43.Cd; 73.61.Wp; 81.05.Uw
Keywords: Graphitic nanocones; Disclinations; Vortex; DiracWeyl equation; Self-adjoint extension; Density of states
1. Introduction
A synthesis of strictly two-dimensional crystals composed of carbon atoms (graphene) [1] is
promising a wealth of new phenomena and possible applications in technology and industry [2].
The observation of anomalous transport properties, and, most exciting, the recent discovery of
substantial field effect and magnetism at room temperature allows one to envisage graphene as a
reasonable replacement of nanotubes in electronic applications, see, e.g., Refs. [35].
* Corresponding author at: Bogolyubov Institute for Theoretical Physics, National Academy of Sciences, 03680 Kyiv,
Ukraine.
E-mail address: yusitenko@bitp.kiev.ua (Yu.A. Sitenko).
doi:10.1016/j.nuclphysb.2007.06.001
242
Yu.A. Sitenko, N.D. Vlasii / Nuclear Physics B 787 [FS] (2007) 241259
By symmetry, the valence and conduction bands in graphene touch at the corners of the hexagonal Brillouin zone. In the vicinity of these points, the dispersion relation is isotropic and linear,
and the density of states at the Fermi level is strictly zero, rising linearly in energy. An effective
long-wavelength description of these electronic states can be written in terms of a continuum
model which is based on the DiracWeyl equation for massless electrons in (2 + 1)-dimensional
spacetime [6,7].
Low-dimensional quantum systems of Dirac fermions can possess rather unusual properties
and, since the discovery of the effect of charge fractionalization [8], are generating current interest. Planar Dirac fermions in the background of the AbrikosovNielsenOlesen vortex [9,10]
were studied in Ref. [11], and, recently, the results of this work have been used to consider the influence of the Kekul distortion in the honeycomb lattice on the electronic properties of graphene
[12,13]. The present paper deals with yet another aspect, and our purpose is to elucidate the role
of topological defects in the graphene lattice.
Topological defects appear as a result of removing (inserting) one or several carbon atoms
from (into) the honeycomb lattice without affecting the threefold coordination of other atoms;
appropriately, the lattice surface is warped owing to positive (negative) curvature induced at the
location of a defect. Assuming that the size of a defect is small as compared to the whole size
of the graphene sample, our interest will be in the study of the influence of such a defect on
the electronic properties of graphene. The consideration is based on the continuum model for
long-wavelength electronic excitations, and various types of defects are characterized by just the
number of carbon atoms removed or inserted; actually, the size of a defect is neglected. The
graphene sheet with a defect takes shape of a cone with the value of the apex angle related to the
number of removed atoms. When a defect is encircled, then sublattices, as well as inequivalent
Fermi points, are entwined or left untwined, depending on the type of a defect. This imposes
a condition on the electronic wave function on the graphene sheet: a phase is acquired under a
rotation around a defect. If the phase commutes with the Hamiltonian, then it can be eliminated
by a singular gauge transformation which, on the other hand, introduces a fictitious point vortex
that may be denoted in the following as a pseudomagnetic one. The flux of the pseudomagnetic
vortex is related to the deficit (proficit) angle of the conical surface, i.e., to the number of removed
(inserted) atoms. For certain types of defects the vortex flux takes fractional values in the units
of 2 .
A general theory of planar relativistic fermionic systems in the background of a point magnetic vortex with arbitrary flux was proposed in Refs. [14,15]; in particular, the case of massless
fermions was considered in Refs. [16,17]. If the vortex flux is fractional in the units of the London flux, then an essentially irregular mode appears among the eigenmodes of the one-particle
Hamiltonian, and this in its turn gives rize to the appearance of an additional parameterthe
self-adjoint extension parameter which specifies the boundary condition at the vortex point [18].
The theory allows us to predict the density of states (not local but total) [19] and the ground
state quantum numbers, as well as their local densities [1517]. The aim of the present paper
is to apply this theory to graphene with a topological defect. The previous attempts to consider
the electronic properties of graphene with topological defects in the framework of the continuum
model approach [2022] have led to contradictory and, therefore, unconvincing results. In our
opinion, this is due to the two circumstances: an inadequate choice of the condition for the electronic wave function in the case of the entwinement of sublattices and an inappropriate treatment
of irregular eigenmodes of the one-particle Hamiltonian. These deficiencies will be remedied in
the present study.
243
In Section 2 we review the derivation of the continuum model for planar graphene in order
to specify the notations used. In Section 3 we introduce topological defects in graphene in the
framework of the continuum model, derive the condition for the electronic wave function in the
case of an arbitrary defect, and demonstrate that the defect can be presented as a pseudomagnetic vortex at the apex of a graphitic nanocone. In Section 4 we consider the solution of the
DiracWeyl equation for electronic excitations on a graphene sheet with a topological defect and
find the density of states and the ground state charge. The results are discussed in Section 5. In
Appendix A we show that the case of the three-pentagon defect coincides actually with that of
the absence of defects. In Appendix B the method of a self-adjoint extension is applied to derive
the condition for the irregular solution to the DiracWeyl equation.
2. Continuum model for long-wavelength electronic excitations
Carbon atoms in graphene form a honeycomb lattice. The Bravais lattice is triangular, and the
primitive cell is rhombic and contains two carbon atoms. If one atom is placed at the origin of
the primitive cell, another one is displaced at d = (d, 0), where d is the lattice spacing. Thus,
the honeycomb lattice is composed of two triangular sublattices: sublattice A (black points) is
generated by vectors ri = ni c1 + mi c2 , and
sublattice B (blank
points) is generated by vectors
3
3
3
3
ri = ni c1 + mi c2 + d, where c1 = ( 2 d, 2 d) and c2 = ( 2 d, 2 d) are the basis vectors of the
primitive cell (see Fig. 1(a)), and ni , mi Z (Z is the set of integer numbers).
Each carbon atom in graphene has four valence electrons. As a result of hybridization, three
of them build -orbitals along the lattice links providing for their rigidity, while the fourth one
makes -orbital which is orthogonal to the lattice plane and is responsible for the conductive
properties of graphene. Retaining only nearest neighbour interactions in the tight-binding approximation, quantum-mechanical hopping of an electron on a planar honeycomb lattice is described
with the use of Hamiltonian
H = t
3

iA j =1
a (ri )b(ri + uj ) t
3
b (ri )a(ri + vj ),
(1)
iB j =1
where t is the hopping amplitude, a (ri ) and a(ri ) (b (ri ) and b(ri )) are the creation and destruction operators acting on sublattice A (B ), which obey anticommutation relations

a(ri ), a (ri ) + = b(ri ), b (ri ) + = ii ,
uj (vj ) are the triad of vectors which are directed to the nearest neighbours of an atom belonging
to sublattice A (B ), see Fig. 1(b),

3
3
1
1
u1 = (d, 0),
u2 =
d,
d ,
u3 =
d,
d ,
2
2
2
2

3
3
1
1
v1 = (d, 0),
(2)
v2 = d,
d ,
v3 = d,
d .
2
2
2
2
Using the Fourier transforms of the sublattice operators,

d 2 k ikri
d 2 k ikri
e
a(k),
b(r
)
=
e b(k),
a(ri ) =
i
(2)2
(2)2
244
(a)
(b)
Fig. 1. The planar honeycomb lattice as a composition of two triangular sublattices. The primitive cell with basis vectors
c1 and c2 is depicted in (a), the triads uj and vj connecting different sublattices are depicted in (b).
where is the first Brillouin zone, Hamiltonian (1) is presented as

H=
d 2k
(k)H (k),
(2)2
(3)
where

T
(k)
= a(k),
b(k)
,

(k) = a (k), b (k) ,
and
H =
j =1 e
ikvj
j =1 e
ikuj
(4)
245
Fig. 2. The first Brillouin zone is a hexagon with opposite sides identified, and, therefore, next to neighbouring corners
are equivalent; two inequivalent ones can be chosen as lying on a vertical line.
Solving the eigenvalue problem, H (k)

= E (k),
one gets

3
3

E = t
eikuj
eikvj
j =1
j =1

3
3
3
= t 1 + 4 cos
ky d cos kx d + cos
ky d .
2
2
2
(5)
As follows from Eq. (5), the one-particle energy spectrum consists of two surfaces (E > 0 and
E < 0) which intersect (E = 0) at six conical points that are located at
kx = 0,
ky = 4(3 3d)1 ,
ky = 2(3 3d)1 .
kx = 2(3d)1 ,
(6)
With one electron per site, the negative-energy states (valence band) are filled and the positiveenergy states (conduction band) are empty, so the band structure is at half-filling with the Fermi
level E = 0 corresponding to six isolated points given in Eq. (6). The first Brillouin zone is
a hexagon with corners identified with these Fermi points; among six of them, only two ones
which can be taken as oppositely located are inequivalent, see Fig. 2.
Actually, H (4) has the meaning of the one-particle Hamiltonian in the momentum representation. The low-energy excitations can be studied by taking the continuum limit (d 0)
at any of
two inequivalent Fermi points. Choosing the pair of inequivalent points as K =
(0, 4(3 3d)1 ) and keeping terms of order k K , one gets:

2
3
0
ix y
1
H = lim d 1 H |k=K + = t
(7)
= hv
x y ,
d0
0
2 ix y
246
where v = 32 t h 1 is the Fermi velocity, and 1 and 2 are the off-diagonal Pauli matrices. Combining the contributions from K+ and K , one gets

0
H +
= h v 1 x + 2 y ,
(8)
0 H
where
1 =
2
0
0
2

,
2 =
1
0
0
1

.
(9)
Making the Fourier transformation of Eq. (8), one gets the long-wavelength approximation for
the one-particle Hamiltonian operator

1
2
H = i hv
(10)
x + y ,
which acts on four-component wave functions of the form
= (A+ , B+ , A , B )T ,
(11)
where subscripts A and B correspond to two sublattices and subscripts + and correspond to
two inequivalent Fermi points. Thus, an effective long-wavelength description of charge carriers
in graphene is written in terms of a continuum model which is based on the DiracWeyl equation
for massless electrons in (2 + 1)-dimensional spacetime, with the role of speed of light c played
by Fermi velocity v c/300 [6,7], see also Ref. [23].
In order to complete the Clifford algebra of anticommuting 4 4 matrices, one has to define
0 , 3 , and 5 = i 0 1 2 3 (where = 0 ). It should be noted that Hamiltonian (10)
commutes with generators Tk (k = 1, 2, 3) of the SU(2)-symmetry transformations, [Tk , Tl ] =
iklm Tm , where
i
1
1
T1 = 3 ,
(12)
T2 = 5 ,
T3 = 3 5 ,
2
2
2
and there is an arbitrariness in the choice of the representation of the Clifford algebra, which is
due to a possibility of diagonalizing anyone of Tk . A representation with diagonal 0 , in view of
the block-diagonal form of 1 and 2 , see Eq. (9), corresponds to the choice of diagonal T3 and
may be denoted as the standard planar representation:
3
1
2

0
0
0
1
2
=
i
=
i
,
,
0 =
0 3
0 1
0 2

0 2
0 2
5
=
i
,
.
3 =i
(13)
2 0
2
0
Choosing T2 to be diagonal, one gets the chiral planar representation:

0 1
0 3
0 1
0
1
2
,
=
,
= i
,
=
1 0
1 0
3 0

0 2
1 0
.
,
5 =
3 = i
0 1
2 0
(14)
A rotation by angle in the plane of a graphene sheet is implemented by operator exp(i),

where

1 1 2 1 3
0
= =
(15)
2i
2 0 3
247
is the pseudospin playing here the role of the operator of spin component which is orthogonal to
the plane. The honeycomb lattice is invariant under the rotation by 2 ,
exp(i2) = ,
(16)
but is not invariant under the rotation by ,

exp(i) = 2i,
i.e., under x x and y y. However, if the latter rotation is supplemented by simultaneous
exchange of sublattices, as well as Fermi points, then this combined transformation,
R exp(i) = i (B , A , B+ , A+ )T ,
(17)
is a symmetry one and can be regarded as the parity transformation for graphene. Note that
transformation (17) is implemented by i 3 i 0 3 in the standard planar representation, see
Eq. (13), or by i 0 in the chiral planar representation, see Eq. (14). The explicit form of R is
extracted from Eq. (17):

0
2
,
R=i
(18)
2 0
and it is given by 5 in representation (13) or by 3 5 in representation (14). Note that R is
commuting with (15) and H (10).
3. Topological defects
Topological defects in graphene are disclinations in the honeycomb lattice, resulting from
the substitution of a hexagon by, say, a pentagon or a heptagon; such a disclination warps the
graphene sheet. More generally, a hexagon is substituted by a polygon with 6 Nd sides, where
Nd is an integer which is smaller than 6. Polygons with Nd > 0 (Nd < 0) induce locally positive
(negative) curvature, whereas the graphene sheet is flat away from the defect, as is the conical
surface away from the apex. In the case of nanocones with Nd > 0, the value of Nd is related to
apex angle ,
Nd
=1
,
(19)
2
6
and Nd counts the number of sectors of the value of /3 removed from the graphene sheet, see
Fig. 3(a). If Nd < 0, then Nd counts the number of such sectors inserted into the graphene
sheet. Certainly, polygonal defects with Nd > 1 and Nd < 1 are mathematical abstractions,
as are cones with a pointlike apex. In reality, the defects are smoothed, and Nd > 0 counts
the number of the pentagonal defects which are tightly clustered producing a conical shape;
such nanocones were observed experimentally [24]. Theory predicts also an infinite series of
the saddle-like nanocones with Nd counting the number of the heptagonal defects clustered in
their central regions. However, as it was shown by using molecular-dynamics simulations [25],
in the case of Nd 4, a surface with a polygonal defect is more stable than a similarly shaped
surface containing a multiple number of heptagons; a screw dislocation can be presented as the
Nd limit of a (6 Nd )-gonal defect.
The two-dimensional DiracWeyl Hamiltonian on a curved surface with the squared length

element ds 2 = gjj (r) dr j r j takes form (see, e.g., Ref. [26])

i
j
H = i hv
(20)
(r) j + j (r) ,
2
sin
248
(a)
(b)
Fig. 3. Formation of a topological defect in graphene: (a) one, two, or three sectors of 60 are removed from the lattice,
(b) if one sector is removed, then sites of different sublattices are identified, a and a , b and b , c and c , etc.
where
j

(r), j (r) + = 2g jj (r)I
(21)
and

i
j (r) = k (r) j k (r) jlk (r) l (r) ,
2

1 ln
l
j k (r) = g (r) j gnk (r) + k gnj (r) n gj k (r) .
2
In the case of a conical surface with a pointlike apex, one has
grr = 1,
g = (1 )2 r 2 ,
(22)
(23)
where r and are polar coordinates centred at the apex, and < < 1. The intrinsic curvature
of a cone vanishes at r = 0 and possesses a 2 (r)-singularity at its apex (r = 0); parameter
249
enters the coefficient before this singularity term. Introducing = (1 ), one gets the metric
in the (r, ) coordinates, which is identical to that of a plane, but with in the range 0 < <
2(1 ). Thus, quantity 2 for 0 < < 1 is the deficit angle measuring the magnitude of
the removed sector, and quantity 2 for < < 0 is the proficit angle measuring the
magnitude of the inserted sector. In the case of graphitic nanocones, parameter takes discrete
values:
Nd
.
=
(24)
6
Using Eqs. (21) and (23), one gets
r = 1 ,
= (1 )1 r 1 2 .
(25)
It is straightforward to calculate the nonvanishing Christoffel symbols
= r
= r 1 ,
r
= (1 )2 r,
and get the spin connection

r = 0,
= 2(1 ).
Thus, Hamiltonian (20) on a conical surface takes form

1
2 1
1
H = i hv
r + r (1 ) i .
(26)
In the case of the planar graphene sheet ( = 0), wave function (11) satisfies antiperiodicity
condition, see Eq. (16),
(r, + 2) = (r, ),
(27)
i.e., the wave function is a section of a bundle with spin connection 2.

Let us consider a graphene sheet with a pentagonal disclination (Nd = 1). When circling once
this defect, the two sublattices in the honeycomb structure are exchanged (see Fig. 3(b)), the
two inequivalent Fermi points are exchanged as well. Circling twice this defect is analogous to
circling once a hexagon in the honeycomb lattice without defects. The situation resembles that
of a Mbius strip, where a double full turn is needed to arrive at the same point. Thus, in the
continuum model description of graphene with a pentagonal disclination, wave function (11) has
to obey the Mbius-strip-type condition:
(r, + 2) = iR(r, ),
(28)
where R is given by Eq. (18), and, consequently,

(r, + 4) = (r, ),
(29)
= I . Note that the sign in the right-hand side of Eq. (28) is chosen by convention.
since
In a similar way, one can show that the wave function on a graphene sheet with a heptagonal
disclination (Nd = 1) obeys the Mbius-strip-type condition as well. Moreover, it can be noted
that sublattices (and Fermi points) are entwined in the case of odd Nd and are left untwined in the
case of even Nd . Thus, the condition for the wave function on a graphene sheet with an arbitrary
disclination takes form

(r, + 2) = exp i Nd R (r, ),

(30)
2
R2
250
where the choice of sign in the exponential function agrees with the sign choice in Eq. (28). Our
results remain unchanged if the opposite sign in Eq. (28) and, correspondingly, in the exponential
function in Eq. (30) is chosen.
By performing singular gauge transformation
Nd
R,
4
one gets the wave function obeying condition
= ei ,
(r, + 2) = (r, ),
(31)
(32)
in the meantime, Hamiltonian (26) is transformed to

3
1
2 1
(1 )1 i R i ,
H = ei H ei = i hv
r + r
2
(33)
where Eq. (24) is recalled. We conclude that a topological defect in graphene is presented by a
pseudomagnetic vortex with flux Nd /2 through the apex of a cone with deficit angle Nd /3.
4. Ground state charge
The density of states is defined as
1
Im Tr(H E i0)1 ,
(34)
Tr is the trace of an integro-differential operator in the functional space: Tr O =

where
d 2 r trr|O|r ; tr denotes the trace over spinor indices only. Since the continuum model of
graphene without topological defects corresponds to the use of the free DiracWeyl Hamiltonian
in flat two-dimensional space, see Eq. (10), the density of states is immediately calculable and
found to be proportional to the size of the graphene sheet
(E) =
S|E|
,
h 2 v 2
where S is its area. The ground state charge of the graphene sheet,
(E) =
e
Q=
2
(35)

(36)
dE (E) sgn(E),
is evidently zero, because Eq. (35) is even in energy.

To consider the influence of topological defects in graphene on the density of states in the
framework of the continuum model, we need the complete set of solutions to the DiracWeyl
equation in this case
H = E ,
H
(37)

is given by Eq. (33), and

is the spinor wave function obeying condition (32). In
where
general, the contribution of a topological defect is added to Eq. (35), and, lacking the factor of
area, it seems to be negligible. However, if this contribution contains a piece which is odd in
energy, then the latter yields the nonzero ground state charge, see Eq. (36). In the following our
interest will be in the search of such a piece, and, as we shall see, its emergence is due to the
appearance of an irregular solution to the DiracWeyl equation.
251
Let us make unitary transformation

= U ,
H = U H U 1 ,
where
1
U = U 1 =
2
then
U RU

=
I
0
I
i 2
0
I
i 2
I
(38)

(39)

(40)
and transformed Hamiltonian H acquires a block-diagonal form:

H1
0

H =
,
0 H1
(41)
where

3
1
Hs = h v i 2 r 1 r 1 (1 )1 is + + 3 ,
2
2
s = 1.
(42)
It should be emphasized that the definite sublattice (A or B) and Fermi-point (+ or ) indices

are assigned to the components of (11), while, after performing transformations (31) and (38),
one gets with components mixing up different sublattices and Fermi points. Certainly, the
calculation of functional trace in Eq. (34) does not depend on the representation used, and it is just
a matter of convenience to use a representation with the block-diagonal form of Hamiltonian (41).
Separating the radial and angular variables

r, |E, n ,
(r, ) =
(43)
nZ
where
fn,1 (r)ei(n+ 2 )
1
gn,1 (r)ei(n+ 2 )
,
r, |E, n =
i(n 12 )
f
(r)e
n,1
(44)
gn,1 (r)ei(n 2 )
one rewrites the DiracWeyl equation as the system of equations for the radial functions

0
Dn,s
fn,s (r)
fn,s (r)
=E
,
gn,s (r)
gn,s (r)
Dn,s
0
(45)
where

1
1
Dn,s = hv
r + r (1 ) (sn ) ,

1
1
= hv
Dn,s
r + r (1 ) (sn + 1 2) .
(46)
A pair of linearly independent solutions to Eq. (45) is written in terms of the cylinder functions.
In the case of 12 < 1 (Nd = 3, 4, 5) the condition of regularity at the origin is equivalent to the
condition of square integrability at this point, and this selects a physically reasonable solution.
Thus, the situation is similar to that of = 0 (absence of a defect), resulting in a density of states
252
which is even in energy. In particular, it can be shown that the density of states in the case of
= 12 (Nd = 3) is given by Eq. (35), see Appendix A.
In the case of 0 < < 12 (Nd = 1, 2) and 12 < 0 (Nd = 1, 2, 3) there is a mode,
for which the condition of regularity at the origin is not equivalent to the condition of square
integrability at this point: both linearly independent solutions for this mode are at once irregular
and square integrable at the origin. To be more precise, let us define in this case

s
nc = sgn() 1 ,
(47)
2
and note that solutions to the DiracWeyl equation correspond to the continuous spectrum and,
therefore, obey the orthonormality condition
2

d
dr r(1 )E, n|r, r, |E , n =
2(E E )
nn ,
EE
(48)
where a factor of 2 in the right-hand side of the last relation is due to the existence of two inequivalent Fermi points. Then the complete set of solutions to Eq. (45) is chosen in the following
form: regular modes with sn > snc

1
Jl(1)1 F (kr)
fn,s (r)
=
,
gn,s (r)
2 (1 ) sgn(E)Jl(1)1 +1F (kr)
l = s(n nc ),
(49)
regular modes with sn < snc

1
fn,s (r)
Jl (1)1 +F (kr)
=
,
gn,s (r)
2 (1 ) sgn(E)Jl (1)1 1+F (kr)
l = s(nc n),
and an irregular mode

1
fnc ,s (r)
=
gnc ,s (r)
2 (1 )[1 + sin(2E ) cos(F )]

sin(E )JF (kr) + cos(E )JF (kr)
,
sgn(E)[sin(E )J1F (kr) cos(E )J1+F (kr)]
where k = |E|(h v)1 , J (u) is the Bessel function of order , and

1 1
sgn() + (1 )1 .
F=
2 2
(50)
(51)
(52)
Thus, the requirement of regularity for all modes is in contradiction with the requirement of
completeness for these modes. The problem is to find a condition allowing for irregular at r 0
behaviour of the mode with n = nc , i.e., to fix E in Eq. (51). To solve this problem, first of all one
has to recall the result of Ref. [27], stating that for the partial Dirac Hamiltonian to be essentially
self-adjoint, it is necessary and sufficient that a non-square-integrable (at r 0) solution exist.
Since such a solution does not exist in the case of n = nc , the appropriate partial Hamiltonian is
not essentially self-adjoint. The Weylvon Neumann theory of self-adjoint operators (see, e.g.,
Ref. [28]) is to be employed in order to consider a possibility of the self-adjoint extension for
253
this operator. We show in Appendix B that the self-adjoint extension exists indeed, and the partial
Hamiltonian at n = nc is defined on the domain of functions obeying the condition

limr0 (rMv/h )F fnc ,s (r)
2F 1 (F )
=
2
tan
+
,
(53)
(1 F )
2
4
limr0 (rMv/h )1F gnc ,s (r)
where (u) is the Euler gamma function, M is the parameter of the dimension of mass, and is
the self-adjoint extension parameter. Substituting the asymptotics of Eq. (51) at r 0 (see, e.g.,
Ref. [29]) into Eq. (53), one gets the relation fixing parameter E ,

h k 2F 1

tan(E ) = sgn(E)
(54)
tan
+
.
Mv
2
4
Using the complete set of solutions, it is straightforward to determine the kernel of the resolvent, r, |(H )1 |r , (where is a complex parameter with dimension of energy),
calculate its functional trace, see, e.g., Ref. [19], and find density of states (34). Only the irregular mode contributes to the odd in energy piece of the density of states, which is given by
expression
|E| 2F 1

|E| 12F
2(2F 1) sin(F ) ( Mv
tan( 2 + 4 ) + ( Mv
cot( 2 + 4 )
2)
2)
(E) =
|E|
.
2(2F 1) tan2 ( + ) 2 cos(2F ) + ( |E| )2(12F ) cot2 ( + )
E ( Mv
2)
2
4
2
4
Mv 2
(55)
Inserting Eq. (55) into Eq. (36), we calculate the ground state charge,

Q = e sgn0 (1 2F ) cos ,
(56)
where
sgn0 (u) =

sgn(u), u = 0
.
0,
u=0
The charge, if any, is accumulated in the vicinity of the defect, and its density is given by expression, see Ref. [17],
2 sin(F )
(r) = e 3
(1 )r 2

0
dw w[K1F (w) KF (w)]

h w 2F 1
( rMv
)
tan( 2
h w 12F
+ 4 ) + ( rMv
)
cot( 2 + 4 )
(57)
decreasing as an inverse power at large distances from the defect; here K (u) is the Macdonald
function of order .
In the case of < 12 (Nd = 4, 5, . . .) there are two or more irregular modes, unless
= 1 (Nd = 6). The case of more than one irregular modes will be considered elsewhere,
while in the case of = 1 the irregular mode appears at n = nc with nc = 2s and has the
form

sin(E )J 1 (kr) + cos(E )J 1 (kr)
1
fnc ,s (r)
2
2
=
;
(58)
gnc ,s (r)
2 (1 ) sgn(E)[sin(E )J 1 (kr) cos(E )J 1 (kr)]
2
hence this case corresponds to F = in Eq. (51), yielding the vanishing ground state charge.
Our results are summarized in Table 1.
If the sign in the exponential function in condition (30) is changed to the opposite, then this
corresponds to change F 1 F . Our results remain unchanged, if, in addition, shift
+ is performed.
1
2
254
Table 1
The ground state charge in the case of the existence of the one irregular mode in the set of eigenmodes of the Hamiltonian
Nd
F
Q
1
2
1
5
5
7
1
2
1
3
1
2
e sgn0 (cos )
e sgn0 (cos )
e sgn0 (cos )
5. Discussion
In the present paper we study the electronic properties of the carbon monolayer (graphene)
with disclinations, i.e., (6 Nd )-gonal (Nd = 0) defects inserted in the otherwise perfect twodimensional hexagonal lattice. The effects of the variation of the bond length or the mixing of with -orbitals caused by curvature of the lattice surface are neglected, and our consideration,
focusing on global aspects of coordination of carbon atoms, is based on the long-wavelength
continuum model originating in the tight-binding approximation for the nearest neighbour interactions. Our general conclusion is that the dependence of the electronic properties on the value
of Nd is not monotonic, but rather abruptly discontinuous. For some values of Nd the density
of states is predicted unambiguously by the theory, whereas, otherwise, its theoretical prediction
involves some parameters which should be determined from the experiment.
As it was already noted [20], a defect with odd Nd entwines two sublattices, as well as two
inequivalent Fermi points, and in the present paper we show that the correct condition for the
electronic wave function involves operator R commuting with the Hamiltonian, see Eqs. (30)
and (18). However, much stronger impact on electronic properties might be drawn by the fact that
for certain Nd irregular modes emerge among the eigenmodes of the Hamiltonian. It is instructive
to compare two cases when the density of states remains almost the same as for planar graphene,
but for different reasons. In the case of the three-pentagon defect (Nd = 3), there is entwinement
of sublattices and there is no irregular modes; the density of states is calculated (see Appendix A)
and shown to coincide exactly with that of planar graphene. In the case of the two-pentagon defect
(Nd = 2), there is no entwinement and there is an irregular mode; the density of states is even in
energy (owing to F = 12 , the odd piece vanishes, see Eq. (55)) and almost coincides with that of
planar graphene. Thus, we disprove the controversial assertions that the density of states at the
Fermi level is nonzero either at Nd = 2 [20] or at Nd = 3 [21].
The same unambiguous predictions are obtained for a graphene sheet with the two-heptagon
defect (Nd = 2) and a graphene sheet with a dodecagon or six heptagons (Nd = 6): the
density of states almost coincides with that of planar graphene. Evidently, the ground state charge
is zero in all above cases.
Let us turn now to the cases when our predictions are not unambiguous, since they involve
self-adjoint extension parameter . These cases include graphene sheets with following defects:
one pentagon (Nd = 1), one heptagon (Nd = 1), and three heptagons (Nd = 3). Contrary
to the assertions in Refs. [20,21], the density of states at the Fermi level is characterized by a
divergent, rather than the cusp, behaviour in these cases, see Eq. (55):
6
sgn(E) Mv 2 25
( |E| )
Mv 2
cot( 2 + 4 ),
6 sgn(E) Mv 2 47
7 Mv 2 ( |E| ) tan( 2 + 4 ),
E0
2 sgn(E) Mv 2 23
3 Mv 2 ( |E| ) cot( 2 + 4 ),
(E) =
Nd = 1,
Nd = 1,
Nd = 3.
(59)
255
Actually, there are three possibilities: cos > 0, cos < 0, and cos = 0. The question of
which of the possibilities is realized has to be answered by experimental measurements. First,
the density of states in the vicinity of the Fermi level can be measured directly in scanning tunnel
and transmission electron microscopy. Secondly, the ground state charge can be measured also,
and our prediction, see Eq. (56) or Table 1, is
Q|Nd =1 = Q|Nd =1 = Q|Nd =3 ,
(60)
while the ground state charge density decreases by power law at large distances from the defect,
see Eq. (57),
11
3
( 13
h
10 ) ( 10 ) 1
5
e 9 sin(/5)
2
4
2 ( rMv ) cot( 2 + 4 ),
20
r
(
)
13
3
( 17
h
14 ) ( 14 ) 1
7
e 9 sin(2/7)
(r) =
2
5
2 ( rMv ) tan( 2 + 4 ),
35
r
(
)
Er
7
7
5
e 32 ( 6 )2 ( 6 ) 12 ( h ) 3 cot( + ),
2
4
12
r rMv
(3)
Nd = 1,
Nd = 1,
(61)
Nd = 3.
The results for the ground state charge in the case of cos > 0 agree with the results of numerical
atomistic calculation of the bond network with the use of recursion methods [30]. The pentagonal
defect, as well as the three-heptagon one, is attractive, and the heptagonal defect is repulsive
to electrons. The charge, negative or positive, is accumulated around the defect, and, at large
distances from it, the decrease is the strongest one for a pentagon and the weakest one for three
heptagons.
It should be noted that at cos = 0 and F = 12 scale invariance is broken, and the appearance
of parameter M with dimension of mass evinces this. In general, irregular mode (51) diverges at
the origin as r with < 1. Scale invariance is respected by the condition of minimal irregularity [14,15,31],
"
=
2 (mod 2),
2 (mod 2),
0 < F < 12 ,
1
2
< F < 1,
(62)
which restricts the behaviour of the irregular mode at the origin to r with < 1/2. Thus, both
scale invariance and minimal irregularity favour definitely the choice of cos = 0, when the
density of states and the ground state charge are trivial. It would be inspiring, if the experiment
could prefer other choices.
Acknowledgements
We would like to thank V.P. Gusynin for stimulating discussions. The research was supported
in part by the Swiss National Science Foundation under the SCOPES project No. IB7320110848. Yu.A.S. acknowledges the support of the Program for Fundamental Research of the
Department of Physics and Astronomy of the National Academy of Sciences of Ukraine and INTAS (grant No. 05-1000008-7865). N.D.V. acknowledges the INTAS support through the PhD
Fellowship Grant for Young Scientists (No. 05-109-5333).
256
Appendix A
In the case of = 1/2 Hamiltonian Hs (42) takes form

3 1 3
2
1 1
2is + +
.
Hs = hv
i r r
2 2
The kernel of the resolvent (the Greens function) of the Hamiltonian is presented as

1 i(n+ s )( ) an11 (r, r ) an21 (r, r )
2
e
,
r, |(Hs )1 |r , =
2
an12 (r, r ) an22 (r, r )
nZ
where the radial components satisfy equations

11
an (r, r ) an21 (r, r )
h v(r + r 1 2sn)
h v[r + r 1 (2sn 1)]
an12 (r, r ) an22 (r, r )

11
1
an (r, r ) an12 (r, r )
hv(
r + r 2sn)
=
h v[r + r 1 (2sn 1)]
an21 (r, r ) an22 (r, r )

2
1 0

= (r r )
;
0 1
r
(A.1)
(A.2)
(A.3)
note that a factor before the delta-function is due to (det gjj )1/2 = [(1 )r]1 . The radial
components can be found in the form

ankk (r,r )

=
dp p
h2 v 2 p 2
kk
an,p
(r, r ),
2
(A.4)
where
11
an,p
(r, r ) = 2J2sn1 (pr)J2sn1 (pr ),
12
(r, r ) = 2pJ2sn (pr)J2sn1 (pr ),
an,p
21
(r, r ) = 2pJ2sn1 (pr)J2sn (pr ),
an,p
22
(r, r ) = 2J2sn (pr)J2sn (pr ).
an,p
Putting
(A.5)
= and taking trace of matrix (A.2), one gets
trr, |(Hs )1 |r ,

1 11
=
an (r, r ) + an22 (r, r )
2
nZ

0
dp p
h 2 v 2 p 2 2
nZ

J2sn1 (pr)J2sn1 (pr ) + J2sn (pr)J2sn (pr ) .
(A.6)
The summation is performed with the use of the Neumanns addition theorem (see, e.g.,
Ref. [29]), yielding the expression,
1
trr, |(Hs )
|r , =
dp p
h2 v 2 p 2
J0
2

p(r r ) ,
(A.7)
257
which is badly divergent at r r. To tame the divergence, we define regularized kernel

r, |(Hs )1 exp tHs2 |r ,
11

21 (r, r )
(r, r ) an,p
dp p exp(t h 2 v 2 p 2 ) i(n+ s )( ) an,p
1
2
,
e
=
(A.8)
12 (r, r ) a 22 (r, r )
2
h 2 v 2 p 2 2
an,p
n,p
nZ
0
where t > 0 is the regularization parameter. Now the limit r r can be taken safely, yielding
1
trr, |(Hs )
exp tHs2 |r, =
dp p exp(t h 2 v 2 p 2 )
2 h2 v 2
h 2 v 2 p 2 2

E1 t2 ,
(A.9)
where

E1 (u) =
du u
e
u
is the exponential integral (see Ref. [29]). Note that, actually, we have reiterated the derivation
for the case of the planar graphene sheet ( = 0): the only difference is that in the latter case all
factors of 2 (including those at the order of Bessel functions) in the right-hand sides of Eq. (A.5)
are absent.
Since Eq. (A.9) is independent of r and , the integration over the surface yields a factor of
its area:

S
Tr(Hs )1 exp tHs2 =
(A.10)
E1 t2 .
2
2
2 h v
The divergence of the last quantity in the limit t 0+ does not contribute to the density of
states, Eq. (34). This is due to a specific form of a discontinuity of the exponential integral
at negative real values of its argument, Im E1 (u i0) = i (u > 0). Consequently, we get
finite result (35).
Appendix B
The partial Hamiltonian corresponding to n = nc takes form, see Eqs. (45)(47) and (52),

0
r + r 1 (1 F )
.
h = h v
(B.1)
r r 1 F
0
Let h be defined on the domain of functions 0 (r) that are regular at r = 0. Then its adjoint h
which is defined by relation

0

dr r(1 ) h (r) 0 (r) =

dr r(1 ) (r) h 0 (r)
(B.2)
acts on the domain of functions (r) that are not necessarily regular at r = 0. So the question
is whether the domain of definition of h can be extended, resulting in both the operator and its
258
adjoint being defined on the same domain of functions. To answer this, one has to construct
the eigenspaces of h with complex eigenvalues. They are spanned by the linearly independent
square-integrable solutions corresponding to the pair of purely imaginary eigenvalues,
h (r) = iMv 2 (r),
(B.3)
where Mv 2 is inserted for the dimension reasons. It is straightforward to show that only one pair
of such solutions exists
i

e 4 KF (rMv/h )
1
,
(r) =
(B.4)
N ei 4 K1F (rMv/ h)
where N is a certain normalization factor. Thus, the deficiency index is equal to (1, 1), and,
according to the Weylvon Neumann theory of self-adjoint operators (see Ref. [28]), the selfadjoint extension of operator h is defined on the domain of functions of the form

fnc ,s (r)
(B.5)
= 0 (r) + c + (r) ei (r) ,
gnc ,s (r)
where c is a complex constant and is a real continuous parameter. Using the asymptotics of
the Macdonald function at small values of its argument (see Ref. [29]), we get

sin( 2 + 4 )2F (F )(rMv/h )F
cei 2
fnc ,s (r)
,
=
(B.6)
gnc ,s (r) r0 iN
cos( 2 + 4 )21F (1 F )(rMv/h )1+F
which can be rewritten in the form of Eq. (53).
References
[1] K.S. Novoselov, D. Jiang, F. Schedin, T.J. Booth, V.V. Khotkevich, S.V. Morozov, A.K. Geim, Proc. Natl. Acad.
Sci. USA 102 (2005) 10451.
[2] A.K. Geim, K.S. Novoselov, Nature Mater. 6 (2007) 183.
[3] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, M.I. Katsnelson, I.V. Grigorieva, S.V. Dubonos, A.A. Firsov,
Nature 438 (2005) 197.
[4] Y. Zhang, Y.-W. Tan, H.L. Stormer, P. Kim, Nature 438 (2005) 201.
[5] A.C. Neto, F. Guinea, N.M. Peres, Phys. World 19 (11) (2006) 33.
[6] D.P. DiVincenzo, E.J. Mele, Phys. Rev. B 29 (1984) 1685.
[7] G.W. Semenoff, Phys. Rev. Lett. 53 (1984) 2449.
[8] R. Jackiw, C. Rebbi, Phys. Rev. D 13 (1976) 3398.
[9] A.A. Abrikosov, Sov. Phys. JETP 5 (1957) 1174.
[10] H.B. Nielsen, P. Olesen, Nucl. Phys. B 61 (1973) 45.
[11] R. Jackiw, P. Rossi, Nucl. Phys. B 190 (1981) 681.
[12] C.-Y. Hou, C. Chamon, C. Mudry, Phys. Rev. Lett. 98 (2007) 186809.
[13] R. Jackiw, S.-Y. Pi, Chiral gauge theory for graphene, Phys. Rev. Lett. 98 (2007) 266402, cond-mat/0701760.
[14] Yu.A. Sitenko, Phys. Lett. B 387 (1996) 334.
[15] Yu.A. Sitenko, Phys. At. Nucl. 60 (1997) 2102;
Yu.A. Sitenko, Phys. At. Nucl. 62 (1999) 1084, Erratum.
[16] Yu.A. Sitenko, Mod. Phys. Lett. A 14 (1999) 701;
Yu.A. Sitenko, Phys. Rev. D 60 (1999) 125017.
[17] Yu.A. Sitenko, Ann. Phys. 282 (2000) 167.
[18] P. de Sousa Gerbert, R. Jackiw, Commun. Math. Phys. 124 (1989) 229.
[19] Yu.A. Sitenko, V.M. Gorkavenko, Nucl. Phys. B 679 (2004) 597.
[20] P.E. Lammert, V.H. Crespi, Phys. Rev. Lett. 85 (2000) 5190.
[21] V.A. Osipov, E.A. Kochetov, M. Pudlak, JETP 96 (2003) 140;
R. Pincak, V.A. Osipov, Phys. Lett. A 314 (2003) 315.
[22]
[23]
[24]
[25]
[26]
[27]
[28]
259
A. Cortijo, M.A.H. Vozmediano, Nucl. Phys. B 763 (2007) 293.

J. Gonzalez, F. Guinea, M.A.H. Vozmediano, Nucl. Phys. B 406 (1993) 771.
A. Krishnan, E. Dujardin, M.M.J. Treacy, J. Hugdahl, S. Lynum, T.W. Ebbesen, Nature 388 (1997) 451.
S. Ihara, S. Itoh, K. Akagi, R. Tamura, M. Tsukada, Phys. Rev. B 54 (1996) 14713.
N.D. Birrel, P.C.W. Davies, Quantum Fields in Curved Space, Cambridge Univ. Press, Cambridge, 1982.
J. Weidmann, Math. Z. 119 (1971) 349.
S. Albeverio, F. Gesztezy, R. Hoegh-Krohn, H. Holden, Solvable Models in Quantum Mechanics, Springer-Verlag,
Berlin, 1988.
[29] M. Abramowitz, I.A. Stegun (Eds.), Handbook of Mathematical Functions, Dover, New York, 1972.
[30] R. Tamura, M. Tsukada, Phys. Rev. B 49 (1994) 7697;
R. Tamura, K. Akagi, M. Tsukada, S. Itoh, S. Ihara, Phys. Rev. B 56 (1997) 1404.
[31] Yu.A. Sitenko, Nucl. Phys. B 342 (1990) 655;
Yu.A. Sitenko, Phys. Lett. B 253 (1991) 138.
Renormalization flow for unrooted forests

on a triangular lattice
Sergio Caracciolo a, , Claudia De Grandi b , Andrea Sportiello a
a Universit degli Studi di Milano, Dipartimento di Fisica and INFN, via Celoria 16, I-20133 Milano, Italy
b Departement of Physics, Boston University, Boston, MA 02215, USA
Received 30 May 2007; accepted 15 June 2007

Abstract
We compute in small temperature expansion the two-loop renormalization constants and the three-loop
coefficient of the -function, that is the first non-universal term, for the -model with O(N ) invariance
on the triangular lattice at N = 1. The partition function of the corresponding Grassmann theory is, for
negative temperature, the generating function of unrooted forests on such a lattice, where the temperature
acts as a chemical potential for the number of trees in the forest. To evaluate Feynman diagrams we extend
the coordinate space method to the triangular lattice.
1. Introduction
Results concerning with graph theory [14], that is properties of a set of points which refer
simply to the notion of adjacency, are of interest in a variety of fields, ranging from pure mathematics to statistical physics and find an enormous amount of applications in natural sciences
besides physics like in biology or in theoretical information science.
Detailed properties of a graph can be derived from the study of the partition function of a qstate Potts model [57] with variables defined on its sites. Indeed this function is strictly related
with the Tutte polynomial of the graph [810] and, for example, the generating polynomial of
spanning trees or unrooted forests on the graph can be recovered by taking the limit q 0.
E-mail addresses: sergio.caracciolo@mi.infn.it (S. Caracciolo), degrandi@bu.edu (C. De Grandi),

andrea.sportiello@mi.infn.it (A. Sportiello).
doi:10.1016/j.nuclphysb.2007.06.012
S. Caracciolo et al. / Nuclear Physics B 787 [FS] (2007) 260282
261
A classical result in algebraic graph theory is Kirchhoffs matrix-tree theorem [11] which
expresses the generating polynomials of spanning trees and rooted spanning forests on a given
graph as determinants associated to the graphs Laplacian matrix. For modern applications see
for example [12,13]. It is quite natural to rewrite these determinants as Gaussian integrals over
Grassmann variables.
Recently [14] it has been shown that the solution of other combinatorial problems on a graph
can be represented in terms of Grassmann integrals, even though non-Gaussian. In particular,
the generating polynomial of unrooted spanning forests on the graph is simply written adding
to a Gaussian term a suitable four-fermion term. Interestingly, the same partition function can
be obtained, order by order in perturbation theory, by considering an anti-ferromagnetic nonlinear -model with O(N) invariance in the limit in which N 1. These representations
are very convenient to study the cases in which the graph is an infinite regular lattice, because
the whole machinery of Statistical Field Theory becomes available. For example, Renormalized
Perturbation Expansion can be used, Renormalization Group notions can be applied and one
sees that on two dimensional lattices these models are asymptotically free [1518]. The same
mapping has been used at the transition at negative tree fugacity which corresponds to the Potts
anti-ferromagnetic critical point [1921].
In this paper we will concentrate on the triangular lattice and, in particular, we are interested in
the evaluation of the so-called -function. We have computed the three-loop coefficient which is
the first non-universal term, which, in contrast with the square lattice, was yet unknown. A direct
practical relevance of this coefficient comes from a recent study of the zeroes of the partition
function of the Potts model in the complex plane of the inverse temperature w by means of the
numerical evaluation of a transfer matrix in a strip [22]. The locum of zeroes converges to a pair
of complex-conjugate curves with horizontal asymptote, but the convergence is very slow in a
region of large Re(w). It turns out that the shape of this curve can be deduced perturbatively (in
1/w) from the expression of the -function, thus in the region where the errors are larger.
2. Unrooted forests
Let G = (V , E) be a finite undirected graph with vertex set V and edge set E. Associate to
each edge e a weight we , which can be a real or complex number or, more generally, a formal
algebraic variable. For i = j , let wij = wj i be the sum of we over all edges e that connect i to j .
The (weighted) Laplacian matrix L for the graph G is then defined by

wij
for i = j,
Lij = w
(1)
for i = j.
k=i
ik
This is a symmetric matrix with all row and column sums equal to zero.
Since L annihilates the vector with all entries 1, its determinant is zero. Kirchhoffs matrixtree theorem [11] and its generalizations [2326] express determinants of square submatrices
of L as generating polynomials of spanning trees or rooted spanning forests in G. For any set of
vertices {i1 , . . . , ir } of V , let L(i1 , . . . , ir ) be the matrix obtained from L by deleting the rows
and columns i1 , . . . , ir . Then Kirchhoffs theorem states that det L(i) is independent of i and
equals

we ,
det L(i) =
(2)
T T eT
262
where the sum runs over all spanning trees T in G. (We recall that a subgraph of G is called a tree
if it is connected and contains no cycles, and is called spanning if its vertex set is exactly V .) The
i-independence of det L(i) expresses, in electrical-circuit language, that it is physically irrelevant
which vertex i is chosen to be ground. There are many different proofs of Kirchhoffs formula (2); one simple proof is based on the CauchyBinet theorem in matrix theory (see e.g. [2]).
The principal-minors matrix-tree theorem reads

we ,
det L(i1 , . . . , ir ) =
(3)
F F (i1 ,...,ir ) eF
where the sum runs over all spanning forests F in G composed of r disjoint trees, each of which
contains exactly one of the root vertices i1 , . . . , ir . This theorem can easily be derived by
applying Kirchhoffs theorem (2) to the graph in which the vertices i1 , . . . , ir are contracted to
a single vertex, while it has theorem (2) as a special case r = 1, through the bijection between
unrooted spanning trees and spanning trees rooted on a given fixed vertex.
Let us now introduce, at each vertex i V , a pair of Grassmann variables i , i . All of these
variables are nilpotent (i2 = i2 = 0), anticommute, and obey the usual rules for Grassmann
integration. Writing

:=
D(, )
(4)
di d i ,
iV
we have, for any matrix A,

A
= det A
D(, )e
and more generally

D(, )
r

(5)
i i eA = det A(i1 , . . . , ir ).
(6)
=1
These formulae allow us to rewrite the matrix-tree theorems in Grassmann form; for instance,
(2) becomes

i i eL
D(, )
(7)
=
we ,
T T eT
while (3) becomes

D(, )
r

i i eL =
(8)
we
F F (i1 ,...,ir ) eF
=1
which is to say

D(, ) exp L + t
i i =
i

F F
F =(F1 ,...,F )

t

i=1

|VFi |
we .
(9)
eF
This formula represents vertex-weighted spanning forests as a massive fermionic free field [2,27].
263
More generally, it has been shown in [14] that

D(, ) exp L + t
wij i i j j
i i + u

F F
F =(F1 ,...,F )
ij
i

t|VFi | + u|EFi |
we
(10)
eF
i=1
where the sum runs over spanning forests F in G with components F1 , . . . , F ; here |VFi | and
|EFi | are, respectively,
the numbers of vertices and edges in the tree Fi . We remark that the
four-fermion term u ij wij i i j j can equivalently be written, using nilpotency of the

Grassmann variables, as (u/2) i,j i i Lij j j . More interestingly, since |VFi | |EFi | = 1
for each tree Fi , we can take u = t and obtain the generating function of unrooted spanning
forests with a weight t for each component.
3. Relation with the lattice -models
Recall that the N -vector model consists of spins i RN , | i | = 1, located at the sites i V ,
with Boltzmann weight eH where

wij ( i j 1)
H = T 1
(11)
ij
and T is the temperature.

Low-temperature perturbation theory is obtained by writing

i = 1 T 2i , T 1/2 i
(12)
with i RN1 and expanding in powers of . Taking into account the Jacobian, the Boltzmann

weight is eH where
H = H +

1
log 1 T 2i
2
(13)

1
T 2 T
Lij i j
i
wij 2i 2j + O 4i , 4j .
2
2
4
i,j
(14)
ij
When N = 1, the bosonic field has 2 components, and so, at least in perturbation theory,
it can be replaced by a fermion pair , if we make the substitution
i j i j i j .
(15)
Higher powers of 2i vanish due to the nilpotency of the Grassmann fields, and we obtain the
model (10) if we identify
t = u = T .
(16)
Note the reversed sign of the coupling: the spanning-forest model with positive weights (t > 0)
corresponds to the anti ferromagnetic N -vector model (T < 0).
264
In the case of a regular unweighted graph of order q, that is all the vertices are connected to
other q vertices, we shall take

1 if i and j are connected,
wij =
(17)
0 otherwise,
and the corresponding Laplacian
1 if i = j and i and j are connected,
if i = j and i and j are not connected,
Lij = 0
(18)
q
if i = j.
This is the case of a regular periodic lattice in d dimensions. If we take unit lattice spacing,
vertices connected to a given site correspond to sites at unit distance, so that, if ek is a lattice
direction, f a lattice function, and x a lattice site, the lattice derivatives are defined as
k f (x) := f (x + ek ) f (x),
(19)
k f (x) := f (x) f (x
(20)
ek ).
The Laplacian can be written as

(Lf )(x) :=
q
k f (x)
(21)
k=1
and when, like in the square and triangular lattice, q is even and to each lattice direction corresponds an inverse lattice direction, that positive directions
(Lf )(x) =
q/2

k k f (x) =
k k f (x) =
k k f (x)
k=1
q/2

q/2
q/2
k=1
k=1

f (x + ek ) + f (x ek ) 2f (x)
(22)
(23)
k=1
and, in the lattice scalar product

f (x)g(x)
(f, g) =
(24)
we have
(g, Lf ) =
q/2

g, k k f
q/2
q/2

k g, k f .
=
(k g, k f ) =
k=1
k=1
(25)
k=1
4. The calculus of the -function

We follow a procedure which already found several applications [2832] for the square lattice.
For a lattice theory, i.e. a theory regularized by introducing a discretization of the coordinates
space, in principle the -function can be found by a direct computation on the lattice, which also
provides a regularization. However, our lattice -model has a natural continuum counterpart,
with the widely investigated action

2 1 ((x) (x))2
1
2
2
h 1 (x) ,
S( , h) = d x
(26)
(x) +
2
2
1 2 (x)
265
where we have introduced an external magnetic field h which explicitly breaks the O(N)invariance. In particular Brzin and Hikami [33] already performed the renormalization up to
three loops in dimensional regularization.
A general theorem of Renormalization states that the n-loop -function within a certain regularization scheme can be deduced from the knowledge of the -function in any other scheme,
at the same perturbative order, and of the renormalization constants in the desired scheme, up to
order n 1. So, a possible procedure, which we will indeed follow in this work, is to relate the
-function on the square and triangular lattice to the continuum results of Brzin and Hikami via
the calculation of the two renormalization constants of the non-linear -model, denoted by Z1
and Z2 .
More in detail, in our case we have to compare our lattice theory with the continuum theory
renormalized in [33] using MS-scheme (Minimal Subtraction modified) and in dimensional regularization. The starting point is the relation for the n-point 1-particle-irreducible (1PI) correlation
functions
n/2
(n)

1/2
(n)
p1 , . . . , pn ; Z11 , Z1 Z2 h;
MS
latt (p1 , . . . , pn ; , h; 1/a) = Z2
(27)
where a and are respectively the lattice spacing and the scale of renormalization for the continuum, while p1 , . . . , pn are the external momenta. Here we consider the lattice theory (denoted by
subscript latt) as a regularization of the continuum theory renormalized at the scale 1/a and we
compare it with the continuum theory renormalized in the MS-scheme (denoted by subscript MS)
at the scale to determine the finite constants Z1 (, a) and Z2 (, a). Both the regularized
theories satisfy a Renormalization Group equation:
d (n)
= 0;
d MS
d (n)
= 0;
da latt
(28)
where we added a minus sign for the lattice equation, because when a 0 we are making a
RG flux toward short distances behaviour, that has the reversed sign respect to the limit
made for the continuum theory. For the lattice theory

n
d (n)
+ W latt () 1 latt ()
0 = a latt = a
da
a
2
1 latt
(n)
latt
() + W () h
,
+
(29)
2
h latt
and analogously for the MS-theory by using W MS () and MS () (in order to avoid confusion
with the coupling constant, and in agreement with the literature on the subject, we denote the function as W ()). By using the condition (27), we are able to join together the and -function
on the lattice to those in MS-scheme. Indeed we find

1 Z1
W MS Z11 = Z1 +
(30)
W latt (),
1

1 Z2 latt
W ().
MS Z11 = latt ()
(31)
Z2 1
The first of them is the important relation that allows us to express the coefficients of the function on the lattice in terms of the coefficients of the continuum theory.
266
Given the -function for the non-linear -model with N the number of vector components,
we expand it in power of the coupling constant 1/ in a generic scheme of regularization
W scheme () =

w0 w1 w2scheme
3
+ O 5 ;
2
4
(32)
the first two coefficients have not the superscript scheme because they are universal, they come
from the calculation respectively at one and two loops (the term from order zero vanishes in two
dimensions); explicitly they are given by
w0 =
N 2
,
2
w1 =
N 2
;
(2)2
(33)
all the other terms are scheme-dependent; the wnscheme coefficient is associated with 1/ n+2 term
of series expansion and correspond to a computation at n + 1 loops. We report here the known
results in MS-scheme (see [33], or [29,30] for other references)
w2MS =
1 N2 4
.
4 (2)3
(34)
We also expand in 1/ the two renormalization constants

(1)
(2)
(0)

Z1
Z
+ 12 + O 3 ,
(0)

Z2
Z
+ 22 + O 3 .
Z1 = Z1 +
(1)
Z2 = Z2 +
(35)
(2)
(36)
With the above conventions on the series expansions, now we look at (30) and we rewrite it as:
W latt () =
W MS (Z11 )
Z1 +
1 Z1
1
(37)
from this equation it can be seen that the coefficient of order n of the expansion of W latt (i.e.
latt ) can be evaluated as long as one knows the coefficients of W MS up the same order (i.e.
wn2
MS ) and performs the computation on the lattice of the constants Z and Z
w1MS , w2MS , . . . , wn2
1
2
up order n 1.1 So we can argue the general result:
latt
latt
= w(n
wn1
-loop) = F

MS
(j )
wi i={0,1,...,n1} ; Z1 j ={0,1,...,n1} .
(38)
For example, for the first scheme-dependent coefficient w2latt , from (37) we find
(1) 2
(2)
(1)
Z1 + w1 Z1 + w2MS .
w2latt = w0 Z1
(39)
1 To be precise, only the constant Z is required. The expansion for Z however comes out as a side result of the
1
2
computation.
267
5. Evaluation of the constants of renormalization

In order to obtain the perturbative expansion of the constants Z1 and Z2 , we use relation (27)
(2)
for the two-point function 1PI. We proceed as follows: we compute latt at n 1 loops and, from
the knowledge of (2) at the same order, and the requirement of validity of (27), we find Z1 and
MS
Z2 at n 1 loops.
For the continuum theory we consider the expansion

(2) (p, , h; ) = p 2
MS
(1)
(p, h; )

+ h + (0) (p, h; ) + MS
+ ;
MS
(40)
we report the already known two-loop results [29,30] in the case of N = 1

1 2
h
p h log 2 ,
4

h
1
h
(1)
2
2
(p, h; ) =
log 2 + 8 log 2 3 + 12(2) R p 2
MS
16 2

h
1
h
2
log 2 + log 2 h
8 2
(0)
(p, h; ) =
MS
(41a)
(41b)
where R is an integral defined as

dpy
dqy
dpx
dqx
1
R := lim h
h0
2
2
2
2 (p 2 + h)(q 2 + h)((p + q)2 + h)

1
1
1
=

,
3
36
24 2
(42)
with (z) = d log (z)/dz, but it appears only in intermediate stages of the computation and
cancels out in any of the results. Therefore

1/2
Z2 (2) p, Z11 , Z1 Z2 h;
MS

(1)

1 2
h
1 (1)
(1)
p h log 2 + Z1 Z2 p 2 Z2 h
= p 2 + h +
4
2

1
h
1
h
1
2 h
2
2
2 h
+
+
8
log
3
+
12(2)
R
p
+
log
log
log
h
16 2
2
2
8 2
2
2

(1)
(1)
(1) 2 Z2(1)
Z
Z
h
(2)
(2)
(1) (1)
+ Z1 Z2 + Z1 Z2 Z1
+
log 2 + 1 2 p 2
4
4
8

(1)
(1)
(1)
(1)
Z2
Z1
Z2
1 (1) 2 1 (2) Z1
h
h
+
Z2
(43)
Z
log 2
log 2
+
h .
8 2
2
4
8
4
8
6. The triangular lattice

On a triangular lattice each site has 6 neighbours. Itis convenient to introduce a redundant
basis of three vectors e(i) , as shown in Fig. 1, such that i e(i) = 0, ei ei = 1, and if i = j then
ei ej = 12 .
268
Fig. 1. Left: the Cartesian basis and the redundant basis on the triangular lattice. Right: Brillouin zone in momentum
space. The rhombus or the hexagon are equivalent choices, as the pairs of triangles denoted with and are related
resp. by periodicity in k1 and k2 . While the hexagon corresponds to the direct construction of the reciprocal lattice, the
rhombus is computationally convenient, as it is a product of one-dimensional intervals.

Lattice sites are labelled by three integers {ni }, with x = i ni e(i) . Because of redundancy,
a constant can be added to the ni s without changing x, i.e. there is an equivalence relation
(n1 , n2 , n3 ) (n1 + m, n2 + m, n3 + m).
(44)
A representative of each class is chosen, for example, by fixing n3 = 0, as

(n1 , n2 , n3 ) (n1 n3 , n2 n3 , 0).
Remark that
x x =

2
3 2 1
ni
ni
.
2
3
i
(45)
(46)

Similarly, the conjugate quantity k = 23 i ki e(i) is characterized by the three numbers ki , such

that i ki = 0. The factor 23 is introduced to have

ki n i .
kx =
(47)
i
As a consequence |ki | < and

kk=
2 2
ki
3
(48)
so that the domain for k is a hexagon of side 2/ 3, or equivalently a (/6)-angle rhombus of

sides 2 (cf. Fig. 1).
Now we can introduce the Fourier transform f(k) for a function f on the triangular lattice

f(k) =
(49)
eikx f (x)
sites
which is such that

f (x) =
hexagon d
2 k eikx f(k)
hexagon d
2k
(50)
269
By specializing this general formula to the gauge (45) we get

f (x) =
dk1
2
dk2 i[k1 (n1 n3 )+k2 (n2 n3 )]

f (k1 , k2 , k1 k2 )
e
2
(51)
where we substituted k3 = k1 k2 and we kept into account the angle of 2/3 between the
vectors e(1) and e(2) in the integration measure. Remark that the volume of the elementary cell
generated by e(1) and e(2) will pop out once more in the continuum limit, indeed

2
d 2 x,
(52)
3
sites

dky
dk1
dk2
dk1
dk2
3
dkx
=
.
(53)
2
2
2
2
2
2
2
7. Tree level
To compare with (26) let us change the normalization of the Grassmann fields to get for the
free part of the action on the triangular lattice

(54)
t
(x) 2(x) (x + ei ) (x ei ) + ht (x)(x)
i
sites
which becomes by Fourier transform

dk1
2

2
dk2
t (k)
k + ht (k)
2
where
k 2 :=
ki2 :=

i

2
ki
=
(2 2 cos ki ).
2 sin
2
(55)
(56)
By using k 2 32 k 2 and (53) this becomes in the continuum limit

2
dkx
2

dky
3 2
t (k)
k + ht (k)
2
2
(57)
and it must be compared with the continuous expression

dkx
2

2
dky
(k)
k + h (k)
2
(58)
from which we get the identifications (see also [34])
t ,
3
3
ht h.
2
(59)
270
In the following it will be useful the evaluation of the integral

I (ht ) :=
dp1
2
1
dp2
2
2 p + ht
(60)
in the limit of small ht . Using the relation

cos p1 + cos p2 = 2 cos
p1 p2
p1 + p2
cos
2
2
(61)
we rewrite the denominator

p 2 + ht = 6 4 cos
p1 p2
p1 + p 2
cos
2 cos(p1 + p2 ) + ht
2
2
(62)
2
2
and then we make the change of variables k1 = p1 +p
and k2 = p1 p
2
2 ; the Jacobian of the
transformation is 2, but it simplifies with the factor 1/2 coming from the new area of integration;
in fact k1 and k2 run inside the rhombus of vertices (, 0), (0, ), (, 0), (0, ), the Brillouin
zone, which is contained twice in the square area [, ]2 . So we obtain

I (ht ) =
dk1
2
1
dk2
.
2 6 4 cos k1 cos k2 2 cos(2k1 ) + ht
(63)
We are now able to integrate in k2 using the result

1
1
d
=
,
2
2 + cos
2
(64)
we have

I (ht ) =

=
2
=
0
1
dk1

2 2 (3 cos(2k ) +
1
ht 2
2)
4 cos2 k1
1
dk1

2 2 ( ht +6 + 2 sin2 k )2 (h + 9)
1
t
2
1
dk1

.
2 2 ( ht +8 cos k )2 (h + 9)
1
t
2
(65)
271
Finally, after the change cos k1 = x, we can express our integral by an elliptic integral2
1
I (ht ) =
2
1
(1 x 2 )( ht 2+8 + ht + 9 x)( ht 2+8 ht + 9 x)

1

1
4 ht + 9
2

=
K
.
h2
2
h2
6 + 2 ht + 9 + 3ht + 4t
6 + 2 ht + 9 + 3ht + 4t
dx
(67)
When ht 0,

ht
1
+ O(ht log ht )
I (ht ) = log
72
4 3
(68)
and therefore, because of (59)

h
1
I (ht ) log
.
48
4 3
(69)
We will also need the evaluation of the integral

I2 (ht ) :=
dp1
2
dp2
1
2
2 (p + ht )2
(70)
in the limit of small ht . Of course

I2 (ht ) =
1
I (ht ) =
+ O(log ht )
ht
4 3ht
(71)
and therefore
1
lim ht I2 (ht ) = .
4 3
(72)
ht 0
Of course the divergent part could be obtained by going to the continuum limit

ht
dp1
2

dpy
dp2
dpx
3 3
1
1
1
h
.
2
2
9
2
2
2 [p + ht ]
2 2
2
2 4 [p + h]
4 3
(73)
2 From 3.148.2 of [35]
u
dx
d
2
1
=
F (, r)
(a x)(b x)(c x)(x d)
(a c)(b d)
(66)

ht +8
ht +8
with a > b > c u > d and = arcsin (ac)(ud)
, r = (ab)(cd)
(cd)(au)
(ac)(bd) . In our case a = 2 + ht + 9, b = 2
ht + 9, c = u = 1, d = 1.

F (, r) = 0 d 2 is the elliptic integral of the second kind, and if = 2 , F ( 2 , r) = K(r) is called the
1r 2 sin
complete integral.
272
Fig. 2. The Feynman diagrams for the two-point function at order 1.
Analogously

dp1 dp2 dq1 dq2

1
2
ht 0
2 2 2 2 (p 2 + h )(q 2 + h )( p
+ q + ht )
t
t
2 2
R
3
dpx dpy dqx dqy
1
2
= ,
lim h
h0
2
3
2 2 2 2 (p 2 + h)(q 2 + h)((p + q)2 + h)
3
(74)
where R was defined in (42).
lim ht
8. One-loop diagrams
The interaction terms on the triangular lattice are

t
+ k)(q)k 2 (p
k)(p)
(p)(p)
(q
2
p
p,q,k
where we introduce the shorthand

dk1
dk2
:=
.
2
2
k
(75)
(76)
We wish to compute the 1PI two-point function. At one loop, two graphs contribute (Fig. 2). On
the triangular lattice, by defining
(k) := k 2 + ht
(77)
we get
2
2 2 1 2 2
p + k 2 i p i ki
p+k
0 (p) = 1
=1
(k)
(k)
k
1
= 1 p I 1 + ht I + p 2 (1 ht I ).
6
By going to the continuum limit, in the limit of small magnetic field
2

p 2 2
p ht I (ht )
6

1
ha 2
2 3 p2 3 2
+ p h log
2
48
3 2 6
4 3
0 (p)
(78)
(79a)
(79b)

1 2
p2
ha 2
= +
p h log
.
48
2 3 4
By comparing the two expressions we obtain the one-loop result

1
3
2 a 2
1
Z1 = 1 +
log
+ +O 2 ,
4
48
2 3

2
2 a 2
1
Z2 = 1 +
log
+O 2
4
48
273
(79c)
(80)
(81)
which, of course, result to be independent from the magnetic field.

9. Two-loop diagrams
The diagrams at second order are the four ones shown in Fig. 3. As we expected these are
the same Feynman diagrams that appear at the second order of perturbative expansion of the
-model [29]. According to the Feynman rules we have to add a minus sign to the diagrams A
and C: for the first one since it has a mass insertion, for the second since it has a loop. So that
the expression of the second order contribution of self-energy is
1 = A B + C D
(82)
2
p+k
,
(k)2
(83)
with
A=
2 2
p+k k+q
B=
,
(q)(k)2
(84)
k,q
Fig. 3. The Feynman diagrams for the two-point function at second order. On the top left corners, we report the identificative letters.
274

C=
k,q

D=
k,q
(q 2 )2
,
(p + q)(k)(k + q)
(85)

p
+ q k 2
.
(q)(k q)(p + k)
(86)
The first two diagrams are easy to evaluate exactly in terms of I and I2 . We find

1
1
A = p 2 I + I2 + ht I2 + I ht I2 ,
6
6

1
1
1
B = p 2 I 2 I + I2 + ht I2 2ht I2 I +
2
3
36
1
1
1
+ 2I 3ht I 2 ht I2 + ht I h2t I2 + 2h2t I2 I.
6
2
6
The diagrams C and D are more involved. First of all remark that

DC=
k,q
k,q
=
k,r
(87)
(88)
(89)
2

k + q k 2 ]
k 2 [p +
k [(p + k + q) (k)]
=
(q)(k + q)(p + k)
(q)(k + q)(p + k)
(90a)
(k)[(p + k + q) (k)]
(q)(k + q)(p + k)
(90b)
(k)[(p + r) (k)]
(r k)(r)(p + k)
(90c)
k,q
where in (90b) we neglect terms of higher order in the small-h expansion. We are interested in
the first terms of the Taylor expansion for small external momentum. We get

sin rj
sin kj
1
4
sin pi sin ki
sin pj
1+
DC
(r k)
(k)
(k)
(r)
k,r

2
(k)
4
p 2
1 2 2
p i ki
sin pi sin ki
+
.
(r) (r) 2(r)
(k)(r)
i
We easily get

1
= I,
(r k)
k,r

(k)
ht
1
= 2I + + ht I 2 I,
(r k)(r)
6
3

(91)
(92)
(93)
k,r
k,r
p 2
= p 2 I 2 ,
(r k)(r)

k,r

1
1
I
2 2
2
p i ki = p +
.
(r k)(r)
3 36
i
(94)
(95)
We have still to compute (changing r into r)

sin rj
sin kj
sin kj
sin ki
+
.
4
sin pi sin pj
(r + k)(k) (k)
(r)
(r)
i,j
275
(96)
k,r
The tensor form of the expression above is

sin pi sin pj ij
(97)
i,j
with ij symmetric under the exchange of i with j , and permutation of indices 1,

2, 3, so that
in general ij = a + bij , which substituted into the previous expression gives a( i sin pi )2 +

b i sin2 pi bp 2 + O(p 4 ) because we have that i pi = 0. Therefore we need only the coefficient b which can be computed, for example, as

sin k1 sin k3 sin k1 sin r1 sin k1
+
.
11 13 = 4
(98)
(r + k)(k) (k)
(r)
(r)
k,r
Then we get

(sin k1 sin k3 ) sin k1
(sin k1 sin k3 ) sin k1
=I
2
(r + k) (k)
2 (k)
k,r
k,r
=I
1
1
1
(I ht I2 )
+
2
12 8 3
and

k,r
k,r
k,r
k,r

R G
I2
1
1
sin2 k1
,
=
I

(r + k)(k)(r)
3
6 2 3
9
4
(99)
sin k1 sin k3
I2
I
G
1
R
= +

,
+
(r + k)(k)(r)
6
18
8
144
4 3
(100)

R
I2
1
1
G
L
sin k1 sin r1
+
= +I

+ + ,
(r + k)(k)(r)
6
12 4 3
18
8
24
(101)
I2
I
G
K
L
1
sin k3 sin r1
R
=

+
(r + k)(k)(r) 12 8 3
36 16 16 48 288
(102)
with
4
k1 + r1 [(r + k) (k) (r)]
G :=
,
2 (r + k)(k)(r)
(103)
k,r

K :=
k,r
k12 k2 k3 r1 k
1 + r1
,
(r + k)(k)(r)
(104)
276

L :=
k,r
2 2
k
1 + r1 k1 r1
.
(r + k)(k)(r)
So finally we found:

3G + K + L
ht
3
1
1
1
2
+R+

I + + ht I 2 I
D C = p I
3 2 3
4
72
6
3
and in conclusion

1
3G + K + L
2I
1
1 = p 2 I 2
+
+R+
+
72 24 3
4
3

1
1
1
+ ht 2I 2 + I

.
+
6 2 3
24 3
By comparing the two expressions (43) and (107) we obtain the two-loop result
2 2
3
9
2 a 2
2 a 2
3
(2)
2 a
Z1 =
log
log
+
log
+
48
4
48
48
16 2
8 2
1
3
3(3G + K + L)
+ ,
+
+
4
16 2 8
2 a 2
2 a 2
5
1
(2)
log2
+ log
Z2 =
2
48
48
16
4 3
(105)
(106)
(107)
(108)
(109)
and the three-loop result by (39)
3
1
3
9(3G + K + L)
latt
w2 =
(110)
+
+
.
2
3
16
8
8
16
By application of the coordinate-space method by Lscher and Weisz [36] suitably modified for
the triangular lattice (see Appendix A) we have obtained the numerical determinations
G = 0.025786368,
(111a)
K = 0.007632210,
(111b)
L = 0.035410394
(111c)
with errors smaller than the quoted digits, from which we recover the value
w2latt = 0.01375000819.
(112)
10. A direct application

The determination of the coefficient w2latt can be used, as shown in Ref. [22], to recover, for
example, the phase boundary in the plane of complex temperature for the q-state Potts model
in the limit q 0. This separatrix is, indeed, a special renormalization-group flow curve. If we
call x and y, respectively, the real and imaginary part of the complex temperature we must have
therefore that

A1 A2
y(x) = y0 1 +
(113)
+ 2 +
x
x
277
where
A1 =
w1
3
w0
1
= ,
2 3
A2 =
w2latt
3 3
w0
2 latt
w
9 2
(114)
and y0 was numerically estimated to be

y0 0.394 0.004.
(115)
For numerical purposes in [22] a variant parametrization is followed, that is

y(x) = y0 exp 1 +
B1
B2
B3
+
+
+
x 0 (x 0 )2 (x 0 )3
(116)
where comparison with (113) in the limit of large x gives the relations
B1 = A1 =
1
,
2 3
B2 = A2
A21
0 A1 .
2
(117)
The parameter 0 , and Ai and therefore Bi with i 2 were not known. In [22] the authors
decided to truncate (116) by setting Bi = 0 for i 3 and try estimated 0 and B2 by this ansatz
by imposing the value of the function and its derivative on the last known numerical point, that
is y(0.0198) = 0.23 and y (0.0198) = 0.369003. They estimated
0 0.550842,
(118)
B2 0.122843.
(119)
From our calculation we get an evaluation of B2 ,

A2 0.00959932,
(120)
0
B2 0.0053776 +
2 3
(121)
so that we can use the strategy just discussed to derive B3 in addition to 0 and B2 . We obtain
0 0.778527,
(122)
B2 0.066160,
(123)
B3 0.162495.
(124)
The curve resulting from this numerical values does not differ substantially from the old one as
can be seen in Fig. 4. This gives more confidence on the method and results in [22].
Acknowledgements
We thank Jesus Salas and Alan Sokal for their interest in our work and for providing us the
numerical information needed to produce Fig. 4.
278
Fig. 4. Phase boundaries for infinite strips of the triangular lattice. Numerical values from different lattice widths L,
from 2 to 9, on gray-tone curves from left to right. Black dots reproduce the extrapolated L limiting curve
in the region of negative Re(1/t). The black dotted-dashed curve and the continuous black curve (almost indistinguishable), in the region of positive Re(1/t), are respectively the old and new curves from the ansatz of Eq. (116).
In the magnification on the right, we plot the discrepancy between the two curves along the Im(1/t) axis, as a function of Re(1/t): as it should, it vanishes with its first derivative, at the numerical point 0.0198 used for the extrapolation,
and vanishes asymptotically because the same estimate of the asymptote y0 is used; all in between, it remains of order 103 .
Appendix A. LscherWeisz method for evaluation of lattice integrals

In the evaluation of two-dimensional lattice integrals, we used the coordinate method illustrated in the paper by Lscher and Weisz [36], and specialized to two dimensions by D.-S. Shin
[31,37], although also the momenta method proposed in Appendix C of [38] could have been
used.
The main idea is the use of some basic relations for the free propagator in coordinate space
(the defining Laplacian equation and a set of relations due to Vohwinkel), in order to find a
recursion which, starting from the values in a certain number of sites neighbouring the origin
(the fundamental lattice integrals), allows to find the whole set of free propagators in lattice sites
in a large radius R, in a time which scales polynomially with R. As a side result, it gives a simple
proof of the fact that all these values are linear combinations with rational coefficients of the
fundamental lattice integrals.
Generalization of the procedure to the triangular lattice is not straightforward, and involves
some delicate points. Some of them are:

In the redundant set of variables (p1 , p2 , p3 ), the constraint i pi = 0 does not allow for
derivatives in a single variable: one should either perform linear combinations of derivatives where the sum of coefficients is zero (for example, (1 3 )f (p1 , p2 , p3 )), or
equivalently, perform derivation within a non-redundant choice of variables (for example,
1 f (p1 , p2 , p1 p2 )).
279
Because of this fact, the Vohwinkel relations involve a larger number of terms, and thus it is
more difficult to manipulate them in order to have a recursion relation. It will turn out that a
larger strip is required for the first x-axis recursion.
For G (x) at values of larger than 1, the choice of subtraction is now not anymore easily
deduced by the Taylor expansion of the exponential and the requirement of periodicity. Now
we also have the requirement of gauge-invariance under x x + m(1, 1, 1), which forces
the application of hat factors only to combinations of pi where the sum of coefficients is
zero.
Coming back to the point, the free subtracted propagator
ikx
e
1
G(x) =
k 2
(A.1)
satisfies the Laplace equation

G(x) = (2) (x)
(A.2)
with the lattice operators

:=
3
i i =
i=1
3

i i
(A.3)
i=1
and we have
1
G(1, 0) = .
6
For the triangular function H (x) defined as

H (x) = eikx ln k 2
G(0, 0) = 0,
(A.4)
(A.5)
a set of Vohwinkel relations holds

G(x + )
G(x i ) G(x + ) + G(x ) = (x x )H (x)
(A.6)
but only two of them (e.g. (, ) = (1, 2) or (1, 3)) are independent.
Using the previous equations and the Laplace equation (A.2), we are able to eliminate H (x),
and write a recursion relation. The one we find on a width-2 strip along the x axis is given by the
set of equations

G (x, 1) ,
0 = 6G(x, 1) +
(A.7)
;

0 = 6G(x, 0) + 2 G(x + 1, 1) + G(x, 1) + G(x + 1, 0) + G(x 1, 0),

0 = G(x 1, 1) G(x + 1, 1) + x G(x, 2) G(x, 0)

+ (x 1) G(x + 1, 2) G(x 1, 0)
(A.8)
(A.9)
which must be solved with respect to G(x + 1, a), with a = 0, 1, 2, in order to have a consistent recursion. A new fundamental integral is required. A choice could be G(2, 1), which is
valued
280
3
1
(A.10)
.
3
In a similar fashion, given the values of G(x) on the width-2 strip, the function can be determined in the whole plane (a sector with x = (n, m), with n 2m 0 is sufficient, because of
symmetry). The Laplacian equation alone is enough
to fulfill this task. So we conclude that at all
values of x the function G(x) is in the set Q + 3 Q.

The integrals of the form
2n
p i
(2n)
K1 =
(A.11)
p 2
G(2, 1) =
which involve G(x) only on the real axis, are easily computed, the first values being
4 4 3
24 3
448 288 3
,
K16 = 16
,
K18 =
+
.
K14 = +
(A.12)
3
The next ingredient we need in order to calculate all the triangular-lattice quantity arising from
our diagrammatics is the two-propagator function in coordinate space. It turns out that the proper
subtraction is the following

G2 (x) =
k
eikx 1 + 14 ((k 2 2k32 )(x1 x2 )2 + cyclics)

.
(k 2 )2
(A.13)
The triangular-lattice Laplacian relation still reads

G2 (x) = G(x)
(A.14)
while the Vohwinkel relation, still for (, ) = (1, 2) or (1, 3), is

1
G2 (x )
G2 (x + ) + G2 (x ) = (x x ) G(x) +
G2 (x + )
4 3
(A.15)
(remark the presence of the corrective contribution 1/(4 3) due to regularization). At the aim
of building the recursion, also in this case it turns out that, as the support of the relations is
identical to the one of the triangular-lattice G(x) case, the independent lattice integrals still must
be the ones located at the points x {(0, 0), (1, 0), (2, 1)}. The first two vanish because of the
subtraction, while the last one is computed analytically, with the result
1
G2 (2, 1) =
4 3
(A.16)
and thus,as it is again a rational times 3/ , still the function G2 (x) at a generic point is in the
set Q + 3 Q.
The two-propagator analogues of the quantities K12n are the integrals of the form
(2n)
K2

=
p
p i2n
.
(p 2 )2
(A.17)
281
They still involve G2 (x) only on the real axis, and thus are easily computed, the first values being
3
1
8 3
176 104 3
4
6
8
,
K2 = 4 +
,
K2 =
.
K2 =
(A.18)
3 3
We need also the lattice sums

2

2
1 1 G(x) G(x) ,
G=
(A.19)
xZ2

K =2
xZ2

1 1 2 + 2 G(x) 3 + 3 G(x) G(x)

2
1 1 2 + 2 3 + 3 G(x) G(x) ,
(A.20)
xZ2
L=

3
1 1 G(x) .
(A.21)
xZ2
At this aim we need the full strength of coordinate method: we evaluate the subtracted propagators G(x) and G2 (x) on lattice points up to a given hexagon of side r ( 25), exactly in terms
of rationals, in negligible computational time (O(r 2 )), from which we deduce the O(r 2 ) largest
terms in the sums above, while the remaining contribution is estimated from the large-distance
behaviour of the integrands. The numerical results are reported in Eqs. (111).
Appendix B. Details on two loop lattice integrals
By using the identity
2[sin + sin + sin ] = ,
(B.1)
valid when + + = 0, in the cases (, , ) = (k1 , k2 , k3 ) or (ki , ri , ki ri ), we easily get

k12 k22 k32
( i sin ki )2
2 3
1
4
=
= 2I 1
6G ,
(r + k)(k)(r)
(r + k)(k)(r)
6
k,r
4
k,r

4
k,r
i,j
k,r
sin ki sin rj
(r + k)(k)(r)
=
k,r

2 3
3K
1
k1 r1 k2 r2 k3 r3
=I 1
+ 3G
+ ,
(r + k)(k)(r)
2
12
[sin k1 + sin r1 sin(k1 + r1 )]2

=
(r + k)(k)(r)

k,r
i=1,2 [sin ki
+ sin ri sin(ki + ri )]
(r + k)(k)(r)
k,r
2 2
k
1 + r1 k1 r1
= L,
(r + k)(k)(r)

L 3K
k1 r1 k
1 + r1 k2 r2 k
2 + r2
=
.
(r + k)(k)(r)
2
2
k,r
We also see that

sin ki + sin ri sin (ki + ri )
2
i

i
ki ri k
i + ri = k1 k2 k3 + r1 r2 r3 k
1 + r1 k
2 + r2 k
3 + r3
282
therefore

k1 k2 k3 i ki ri k
k1 r1 k
1 + r1
i + ri
i ki ri ki + ri
=
= 3K
(r + k)(k)(r)
(r + k)(k)(r)
k,r
(B.2)
k,r
computed either using the first two lines or the last two lines of the previous block of identities.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
H.N.V. Temperley, Graph Theory and Applications, Ellis Horwood Series, 1981.
N. Biggs, Algebraic Graph Theory, second ed., Cambridge Univ. Press, 1993.
C. Godsil, G. Royle, Algebraic Graph Theory, first ed., Springer-Verlag, 2001.
R. Diestel, Graph Theory, Springer-Verlag, 2000.
R.B. Potts, Proc. Cambridge Philos. Soc. 48 (1952) 106.
F.Y. Wu, Rev. Mod. Phys. 54 (1982) 235;
F.Y. Wu, Rev. Mod. Phys. 55 (1983) 315.
F.Y. Wu, J. Appl. Phys. 55 (1984) 2421.
W.T. Tutte, Can. J. Math. 6 (1953) 80.
I.M. Gessel, B.E. Sagan, Electron. J. Combin. 3 (1996) 2.
A.D. Sokal, in: B.S. Webb (Ed.), Surveys in Combinatorics, Cambridge Univ. Press, 2005, p. 173, math.CO/
0503607.
G. Kirchhoff, Ann. Phys. Chem. 72 (1847) 497.
R. Shrock, F.Y. Wu, J. Phys. A 33 (2000) 3881, cond-mat/0004341.
M.L. Glasser, F.Y. Wu, Ramanujian J. 10 (2005) 205, cond-mat/0309198.
S. Caracciolo, J.L. Jacobsen, H. Saleur, A.D. Sokal, A. Sportiello, Phys. Rev. Lett. 93 (2004) 080601, cond-mat/
0403271.
A.M. Polyakov, Phys. Lett. B 59 (1975) 79.
E. Brzin, J. Zinn-Justin, J.C. Le Guillou, Phys. Rev. D 14 (1976) 2615.
E. Brzin, J. Zinn-Justin, Phys. Rev. B 14 (1976) 3110.
W.A. Bardeen, B.W. Lee, R.E. Shrock, Phys. Rev. D 14 (1976) 985.
J.L. Jacobsen, H. Saleur, Nucl. Phys. B 716 (2005) 439, cond-mat/0502052.
J.L. Jacobsen, H. Saleur, Nucl. Phys. B 743 (2006) 207, cond-mat/0512058.
Y. Ikhlef, J.L. Jacobsen, H. Saleur, A staggered six-vertex model with non-compact continuum limit, cond-mat/
0612037.
J.-L. Jacobsen, J. Salas, A.D. Sokal, J. Stat. Phys. 119 (2005) 1153, cond-mat/0401026.
S. Chaiken, SIAM J. Alg. Disc. Meth. 3 (1982) 319.
J.W. Moon, Discrete Math. 124 (1994) 163.
A. Abdesselam, Adv. Appl. Math. 33 (2004) 51, math.CO/0306396.
S. Caracciolo, A.D. Sokal, A. Sportiello, Grassmann integral representation for spanning hyperforests, arXiv: 0706.
1509.
B. Duplantier, F. David, J. Stat. Phys. 51 (1988) 327.
M. Falcioni, A. Treves, Nucl. Phys. B 265 (1986) 671.
S. Caracciolo, A. Pelissetto, Nucl. Phys. B 420 (1994) 141, hep-lat/9401015.
S. Caracciolo, A. Pelissetto, Nucl. Phys. B 455 (1995) 619, hep-lat/9510015.
D.-S. Shin, Nucl. Phys. B 546 (1999) 669, hep-lat/9810025.
B. Alls, S. Caracciolo, A. Pelissetto, M. Pepe, Nucl. Phys. B 562 (1999) 581, hep-lat/9906014.
E. Brzin, S. Hikami, J. Phys. A 11 (1976) 1141.
S. Caracciolo, Nucl. Phys. B 180 (1981) 405.
I.S. Gradshteyn, I.M. Ryzhik, Table of Integrals, Series, and Products, seventh ed., Academic Press, 2007.
M. Lscher, P. Weisz, Nucl. Phys. B 445 (1995) 429, hep-lat/9502017.
D.-S. Shin, Nucl. Phys. B 525 (1998) 457, hep-lat/9706014.
S. Caracciolo, A. Pelissetto, P. Menotti, Nucl. Phys. B 375 (1992) 195.
Integrable spinboson models descending from rational

six-vertex models
L. Amico a, , H. Frahm b , A. Osterloh b , G.A.P. Ribeiro c,d
a MATIS-INFM & Dipartimento di Metodologie Fisiche e Chimiche (DMFCI), Universit di Catania,
viale A. Doria 6, I-95125 Catania, Italy

b Institut fr Theoretische Physik, Leibniz Universitt Hannover, Appelstrae 2, D-30167 Hannover, Germany
c Theoretische Physik, Universitt Wuppertal, Gaussstrae 20, D-42097 Wuppertal, Germany
d Universidade Federal de So Carlos, Departamento de Fisica, CP 676, 13565-905 So Carlos-SP, Brazil
Received 24 March 2007; accepted 13 July 2007

Abstract
We construct commuting transfer matrices for models describing the interaction between a single quantum spin and a single bosonic mode using the quantum inverse scattering framework. The transfer matrices
are obtained from certain inhomogeneous rational vertex models combining bosonic and spin representations of SU(2), subject to non-diagonal toroidal and open boundary conditions. Only open boundary
conditions are found to lead to integrable Hamiltonians combining both rotating and counter-rotating terms
in the interaction. If the boundary matrices can be brought to triangular form simultaneously, the spectrum
of the model can be obtained by means of the algebraic Bethe ansatz after a suitable gauge transformation;
the corresponding Hamiltonians are found to be non-Hermitian. Alternatively, a certain quasi-classical limit
of the transfer matrix is considered where Hermitian Hamiltonians are obtained as members of a family of
commuting operators; their diagonalization, however, remains an unsolved problem.
1. Introduction
Models describing the interaction between bosonic modes and spin degrees of freedom play
an important role in physics. The prototype application is provided by atomradiation interactions [1], where the atomic dipole is described by an effective spin interacting with a single
E-mail address: lamico@dmfci.unict.it (L. Amico).

doi:10.1016/j.nuclphysb.2007.07.022
284
L. Amico et al. / Nuclear Physics B 787 [FS] (2007) 283300
mode electric field [2,3]. More recently, this class of models has been considered to investigate
various aspects in mesoscopic physics where a given two-level system interacts with a bosonic
environment [4]. Furthermore, we mention applications to analyze the decoherence in superconducting circuits [5,6], and to describe ions in harmonic traps [7]. Both these quantum devices
are of potential importance for quantum computation [8,9]. Finally, applications have emerged
in quasi-2D semiconductors in transverse magnetic field [1012].
In this paper we focus on models describing a single spin S interacting with a single mode boson field which were introduced originally by Jaynes, Tavis and Cummings [2,3]. In two cases the
corresponding model Hamiltonian is known to be exactly diagonalizable with elementary means:
for S where the model is quadratic [13] and for weak or resonant interactions, where the
Rotating Wave Approximation (RWA) can be applied [2,3,14]. In the traditional framework of
laser physics, the RWA assumes that the relevant dynamics is given by coherent oscillations of
the population of the atomic energy levels. Therefore only photon emissions accompanied by
an excitation of the atom (and vice versa), described by operators of the type a S , are taken
into account in the atomic dipole coupling with electric field E d (a + a )S x . The so-called
counter-rotating terms a S + + aS , although contained in the coupling, can be neglected at resonant cavity frequency 0 since they induce a short time dynamics (compared to the rotating
terms aS + + a S ); as a result of the approximation the operator S z + a a is conserved. Such a
picture is robust for weak fields and as long as the cavity mode can be tuned conveniently. In the
most recent applications, however, one or even both these conditions are not met. In superconducting circuits, for example, the effective spinboson coupling to the electromagnetic field of
the device can be naturally large, thereby involving multi-boson processes, which are characteristic for the presence of counter rotating terms [15]; furthermore numerical inspections indicate
that such terms can bend the energy levels whose slope should be controlled reliably to study how
the noise affects the coherence in the system [16]. Finally we mention that in semiconductors,
rotating and counter rotating terms result from Rashba and Dresselhaus spinorbit interactions;
the physical origin of the spin transport in such systems is an open problem that is intensively
studied [12].
From the mathematical point of view, for finite spin representations and generic spinboson
interaction the Hamiltonian is a non-trivial matrix of infinite rank. The main complication comes
from the absence of a conserved number operator of the type S z +a a as in the RWA that would
have made the problem finite dimensionala a is the bosonic number operator. In this sense,
using the language of spin chains, the problem will have the complexity of an XYZ chain. We
will speak of a spinboson model to contain rotating terms only, if the Hamiltonian action only
connects a subspace of finite dimension. It is worth noting that it is the simultaneous presence
rotating, S a, and counter-rotating terms S a that makes the Hamiltonian non-trivial in the
sense that a reduction to a finite dimensional subspace is impossible.
In this work we study integrable spinboson models that can be constructed by means of the
quantum inverse scattering method (QISM) [17]. The central object is the transfer matrix, that
is a generating function of integrals of the motion, including the Hamiltonian by definition. The
formalism includes a certain freedom to choose (twisted) periodic boundary conditions and open
boundary conditions, where two boundary matrices can be included, representing boundary magnetic fields in the standard formalism for spin chains [18,19]. Twisted boundary conditions have
been investigated for XXX-symmetry, i.e. for models emerging from the XXX R-matrix in [22]
and Hermitian spinboson models have been created that still have a conserved number operator,
indicating that the apparent counter-rotating terms are spurious. In the present work, open boundary conditions for XXX-type symmetry are analyzed, where two non-parallel boundary matrices
285
are meant to break the XXX-symmetry completely. Integrable open boundaries, corresponding to
dynamic boundary matrices (i.e. depending on the spectral parameter) obey the reflection equation due to Sklyanin [19]. Two approaches are used. First, a bosonic Lax matrix, relative to the
XXX R-matrix is employed and the QISM and algebraic Bethe ansatz is elaborated for this case.
Second, the spinboson Hamiltonian is obtained from a two-site spin chain with mixed representations by virtue of a large S limit (algebraic contraction) of one of the spins to a bosonic
degree of freedom.
The paper is laid out as follows. In the next section we summarize the ingredients of the reflection equations to deal with integrable models with generic boundaries. In Section 3 we present
the spinboson transfer matrix. The diagonalization will be pursued by a direct algebraic Bethe
ansatz (Section 4.1) and by means of the algebraic contraction of a suitable auxiliary spin problem (Section 4.2). The Hamiltonian models extracted from the transfer matrix will be presented
in Section 5. Section 6 is devoted to the conclusions.
2. Integrable boundary conditions
In this section, we review the basic aspects of the quantum inverse scattering method providing
integrable theories with non-trivial boundary conditions.
The central object of the method is the transfer matrix t (), which can be conveniently written
as the trace over an auxiliary space A of an ordered product of L distinct Lax matrices Lj ()
acting in the site j . It can be explicitly written as

t () = trA T () , T () = LL ()LL1 () L1 (),
(1)
where T () is called monodromy matrix.
The R-matrix and Lax matrices Lj () are solutions of the YangBaxter equations
(2)
(2) (1)

R12 ( )L(1)
j ()Lj ( ) = Lj ( )Lj ()R12 ( ),

R12 ()R13 ( + )R23 ( ) = R23 ( )R13 ( + )R12 (),
(2)
(3)
(m)
Lj
indicates that in a multiple tensor product of auxiliary spaces, the

where the suffix m in
Lax matrix acts non-trivially only on the mth copy. A rational solution of (3) is the following
R-matrix
+ 0 0
0

0
0
R() =
(4)
,
0

0
0
0 0 +
where is a C-number named quasi-classical parameter. The YangBaxter equation for the Lax
matrices (2) provides the condition to the commutation property of the transfer matrix (1) and
give rise to the fundamental commutation relations (FCR) for their operator entries.
In the realm of the QISM boundary terms can be considered. The simplest way to do it is to
construct the monodromy matrix in terms of modified Lax matrices L() GL(), G being a
matrix of constants (with respect to the spectral parameter) [18]. The corresponding theory is
still integrable so far the quadratic relation (2) is satisfied. For that it suffices that

R12 (), G (1) G (2) = 0.
(5)
For the case of R-matrix (4) this relation is satisfied by any (2 2)-matrix with entries g11 , g12 ,
g21 , g22 independent of the spectral parameter.
286
It turns out that such modified monodromy matrices correspond to theories with toroidal
boundary conditions. In the case where the boundary matrix G is diagonal the corresponding
transfer matrix (1) can be diagonalized with very little difference from the periodic case because
it does not change in a drastic way the properties of the monodromy matrix elements. In general
this does not occur when G is non-diagonal. Nevertheless, for the rational solutions of the Yang
Baxter equation (3), it is possible to apply a combination of suitable auxiliary and quantum space
transformations which relate the non-diagonal boundary problem with a diagonal one [27]. These
transformations are the irreducible representations of the symmetry of the auxiliary and quantum
space of the L-operator Lj (). Such invariance of the Lax matrix can be expressed by [28]
1 (1) (1)

(1)
Lj () G Uj ,
Lj () = G (1) Uj
(6)
where Uj stand for the irreducible representation of the symmetry of the quantum space [27].
For instance, if the quantum space is invariant by spin-S representation of SU(2), the relation (6)
is satisfied by
u
u
S,S
S,S1
uS1,S
Uj =
..
.
uS,S
uS1,S1
..
.
uS,S1
..
.
S,S
uS1,S
,
..
.
uS,S
where some of the matrix elements ui,j are
(2S)!
2Sj k k
uS,Sk = g1,1 g1,2
,
k!(2S k)!
(7)
(8)
while the remaining satisfy the following recurrence relation
(g1,1 g2,2 g1,2 g2,1 ) n(2S n + 1)

uSk1,Sn =
uSk,Sn+1
2
(k + 1)(2S k)
g1,1

g2,1 2 k(2S k + 1)
+
uSk+1,Sn
g1,1
(k + 1)(2S k)
+
2(S k)
g2,1
uSk,Sn ,
g1,1 (k + 1)(2S k)
(9)
and k, n are integers satisfying k, n = 0, . . . , 2S.

As the transfer matrix (1) gives origin to a commutative family of operators, it can be regarded
as generating functional of the integrals of the motion of the quantum theory, among which one of
them can be taken as the Hamiltonian. For example the logarithmic derivative of t () provided a
protocol to demonstrate the integrability of a variety of local models (that is with nearest neighbor
interactions) [17]. For later use we remark that an arbitrary function of t () or e.g. its coefficients
in a Taylor series in can be taken to be the Hamiltonian.
A different class of lattice models can be obtained focusing on the quasi-classical parameter .
Specifically, integrable long-range Hamiltonians, as the Gaudin and the BCS model, emerge from
the so-called quasi-classical limit 0 of vertex models [2326].
More complicated boundary terms can be considered in the integrable theory via a procedure
demonstrated by Sklyanin [19,30]. To begin with, let us impose the following constrains on the
287
R-matrix:
Unitarity:
R12 ()R12 () = ()1;
(10)
Parity invariance:
P12 R12 ()P12 = R12 ();
(11)
Temporal invariance: R12 ()
(12)
Crossing symmetry:
= R12 ();
t1
t1
R12 ()R12 ( 2) = ( + )1;
(13)
t1 t2
where () = 2 2 and the crossing parameter = . Here 1 is the identity matrix, P12 is the
permutation operator, t denotes transposition on the th space.
According to the Sklyanin procedure, open boundary conditions preserve the integrability of
the model need to provide 2-particle scattering in the bulk compatible with the scattering off the
boundaries. Such boundary conditions can be parametrized by boundary matrices K() which
satisfy the so-called reflection equations
1
R12 ( )K()R21 ( + )K( ) = K( )R12 ( + )K()R21 ( ),

1
t
t
()R12 ( 2)K+
( )
R21 ( + )K+
2
t
t
( )R21 ( 2)K+
()R12 ( + ),
= K+
1
(14)
where K () = K() 1, K () = 1 K(), K t is the transpose of K, and R12 is a solution of

the YangBaxter equation (3).
General non-diagonal C-number solutions in case of the rational R-matrix (4) are the K-matrices [31]

+
K () =
(15)
.

Within this framework the analogue of the row-to-row transfer matrix (1) as generator of
commuting integrals of motion in the case of open boundaries is the following operator [19]

1

t () = trA K+ ( + ) T () K ()T () .
(16)
often called double row transfer matrix in the literature. For diagonal K-matrices the diagonalization of t () can be done by an extension of the algebraic Bethe ansatz [19]. For models
based on the rational R-matrices (4) considered in this paper this approach can be extended to
certain non-diagonal K-matrices by combination of suitable transformations in both quantum
and auxiliary space [32,33].
3. Quantum inverse scattering for single mode spinboson model
In this section we apply the general theory presented in the previous section to the spinboson
problem. Starting from the R-matrix (4) we consider two types of Lax matrices satisfying (2):
the spin degree of freedom is described by

Sj
+ + Sjz
(s)
Lj () =
(17)
,
Sj+
+ Sjz
where the operators Sjz and Sj are chosen to be irreducible representations of su(2). On way to
construct a Lax matrix for the bosonic degree of freedom is using a representations of su(2) in
288
terms of Bose operators in (17), e.g. the HolsteinPrimakoff or the DysonMaleev representation.
Here, however, we choose the bosonic Lax matrix as [17,37]

aj
aj aj aj
(b,1)
(b,2)
;
Lj () =
Lj () =
(18)
aj

aj aj aj + +
with [aj , ak ] = j k . For the construction of the transfer matrix (16) we also need the inverse of
the bosonic Lax operator which reads

aj
(b,1) 1
(b,1) 1 (b,2)
(b,1)1
Lj
= detq Lj
() = detq Lj
Lj ().
aj aj aj +
(19)
(b,1)
(b,1)
Here detq [Lj ] = (/) is the quantum determinant of Lj .

The R-matrix (4) intertwines both the spin and bosonic Lax matrices; additionally it is GL(2)
invariant, namely it commutes with G G for all G GL(2) (see (5)). This GL(2) invariance
of the R-matrix implies that GL(2)-transformed Lax matrices G1 LG2 are again solutions of the
YangBaxter relation (2). At a formal level, GL(2) transformations preserves the structure of the
spin Lax matrix (17) and essentially correspond to a rotation of the local spin. We point out that
this is not the case for the bosonic Lax matrix (18); though general GL(2) transformations lead to
canonical transformations of the boson degrees of freedom, only diagonal GL(2) transformations
preserve the structure of the Lax operator.
In what follows we restrict ourselves to setups with a single spin and a single bosonic mode,
corresponding to a two-site monodromy matrix. Many relevant features are already contained
in this simple case, and an extension to the multi-spin and multi-mode situation is technically
straight forward. Integrable spinboson models can be constructed e.g. from the transfer matrix
for toroidal boundary conditions (periodic boundary with two boundary twists)

ttwist () = trA Gs L(s) ( z0 )Gb L(b) ( z1 )
(20)
where Gs , Gb GL(2) do not depend on ; the quantities z0 , z1 , shifting the spectral parameter locally, are known as impurities in the jargon of QISM. Since [ttwist (), ttwist ()] = 0, in
particular all coefficients of a -expansion of ttwist () commute with each other. The first order coefficient of this -expansion of the transfer matrix is a linear combination of the spin and
boson operators occurring in the Lax matrix. This operator, where spin and boson degrees of
freedom are decoupled, consequently commutes with every Hamiltonian constructible from this
transfer matrix. This implies that the problem can be effectively reduced to a dressed spin degree
of freedom, in which the Hamiltonian is block-diagonal [1]. This can be argued from the GL(2)
symmetry of the R-matrix discussed above, in that -independent twists are elements of GL(2)
and can hence be absorbed in a redefined spin and boson Lax operator. Note that this result is
independent of the specific bosonic Lax matrix chosen; either choice of a bosonic Lax operator,
as given in Eq. (18) or obtained from the spin Lax operator (17) inserting e.g. the Holstein
Primakoff or DysonMaleev bosonic representation of su(2), will produce model Hamiltonians
with a conserved number operator S z + a a. Consequently all Hamiltonians constructed in this
way are of block diagonal form, with more or less complicated non-linearity in the bosonic
and the interaction part. Examples for such non-linear generalizations of Hs -ph can be found
in [3640], where both twist matrices were chosen as the same diagonal matrix. Non-diagonal
Hermitian twist matrices have been used in [22].
For obtaining a model including counter rotating terms from the XXX symmetric R-matrix,
we look at open boundary conditions. This amounts to using the transfer matrix (16) as the
289
generator of integrable Hamiltonians. To this end we will consider the following monodromy
matrix
1
T () = K+ ( + )L(s) ( z0 )L(b) ( z1 )K ()L(b) ( z1 )L(s) ( z0 )

(21)
where the boundary matrices are taken from Eq. (15).
In the next section we discuss the exact solution of the eigenvalue problem for the transfer
matrix t () = tr T () by means of the algebraic Bethe ansatz following two approaches. First
by direct algebraic Bethe ansatz, i.e. including the bosonic Lax operator (18), and then by first
solving an auxiliary spinspin problem and then performing a suitable large-spin limit leading to
the bosonic degree of freedom. The explicit construction of model Hamiltonians from (21) will
be presented in Section 5.
4. Algebraic Bethe ansatz for the spinboson transfer matrix
The starting point of the exact diagonalization of the transfer matrix by the algebraic Bethe
ansatz is to identify a suitable eigenstate of the transfer matrix, the so-called pseudo-vacuum |.
In the presence of non-diagonal boundary matrices, this state can be identified by a method proposed in Refs. [27,32]. Sketching their approach, the boundary matrices are brought to triangular
form by a suitable similarity
x r transformation of the monodromy matrix (21). This amounts to
finding matrices Mj = yjj sjj in
T () = M01 T ()M0
1
= M01 K+ ( + )M5 M51 L(s) ( z0 )M4 M41 L(b) ( z1 )

M3 M31 K ()M2 M21 L(b) ( z1 )M1 M11 L(s) ( z0 )M0
(22)
such that the lower left entry of both M01 K+ ( + )M5 and M31 K ()M2 vanishes. The remaining parameters in the Mj have to be chosen such that the lower left entries of the transformed
Lax operators L(s,b) annihilate the spin and bosonic vacuum states |s and |b , respectively. As a result we can choose the product state | = |s |b as pseudo vacuum for
the algebraic Bethe ansatz. Following Refs. [27,32] for the spin Lax matrix we readily obtain
x0 /y0 = x1 /y1 and x4 /y4 = x5 /y5 . Note that having Mj1 L() Mj 1 | triangular implies the
1
same for M j 1 L() ( z0 )M j | , where the matrices M j may differ from Mj in their right
column only. Since the transfer matrix is independent of this spurious freedom, we are free to
choose M3 = M2 , M4 = M1 , and M5 = M0 without loss of generality. If all Lax matrices are
spin Lax matrices, all Mj can be chosen to be equal [27,32], corresponding to a global rotation
of the spin.
4.1. Algebraic Bethe ansatz with the bosonic Lax operator
In this section, we adapt this procedure to the spinboson model for the two-site monodromy (21) and the choice L(b) = L(b,1) ; the alternative choice L(b) = L(b,2) is discussed below.
The approach described next for open boundary conditions equally applies to toroidal boundary
conditions.
290
The lower left entry of the transformed bosonic Lax matrix L (b) () M21 L(b) ()M1 is (up
to a rescaling of the spectral parameter)

x1 y2 +
x2 + y2 a (x1 a y1 ).
(23)
Clearly, this operator can annihilate a vacuum state for all only if x1 y2 = 0. The bosonic
creation operator a has no right eigenstate, therefore we have to choose y2 = 0. This means
that the gauge matrix left to the bosonic Lax matrix must already be in upper triangular form.
Consequently, the pseudo-vacuum of the bosonic degree of freedom is the coherent state (x1 a
y1 )|b = 0 with
x0 x 1 p
=
e .
y0
y1
(24)
We also comment that for Eq. (23) the structure of two reference states found in [41] is lost in
the spinboson case.
(b) (b)
Writing L (b) = A (b) B (b) we have
C
C (b) |b = 0,

y1
( z1 )ep |b ,
A (b) |b =
x2
r1 ep s1
D (b) |b =
|b .
s2
(25)
The characteristic factor ep is obtained from the condition that K + = M01 K+ M 0 be upper triangular
ep =
K+;21
+ K+;22
(26)
where + is an eigenvalue of K+ .
Note that the parameters r0 , s0 , r0 , s0 as well as r2 , s2 , r2 , s2 can be chosen such that both
boundary matrices K+ and K are diagonal. These parameters enter both the Bethe equations
and the expression for the eigenvalue () of the transfer matrix, Eq. (30). The transfer matrix
and its spectrum, nevertheless, is independent of this choice. This freedom provides with a variety
of isospectral Hamiltonians and corresponding Bethe equations. The transfer matrix, however,
differs from that obtained from Eq. (21) and corresponding diagonalized K+ and K . This is
due to the fact that for rj , sj not being all the same, the gauge transformations of the Lax
operators do not induce canonical transformations on the spin and bosonic degrees of freedom.
Here, rather than bringing the boundary matrices into diagonal form, we choose gauge transformations to simplify the Bethe ansatz equations. To be specific, let det Mj = xj sj yj rj = 1
and xj = rj = 1 for j = 0, 1 giving sj = yj + 1 = ep + 1. Furthermore, the condition y2 = 0 for
the existence of the bosonic pseudo vacuum implies that the boundary matrix K in (22) must
be upper triangular (when bringing the monodromy matrix in lower triangular form instead, the
particular role of xj and yj is taken by rj and sj , respectively). Therefore, we are free to choose
M2 1 and the action (25) of L (b) ( z1 ) on the bosonic coherent state simplifies
291
C (b) |b = 0,
A (b) |b = ( z1 )|b ,
D (b) |b = |b .
(27)
Similarly, after the similarity transformation, the action of the spin Lax operator L(s) ( z0 ) for
a 2S + 1 dimensional representation of su(2) on |s becomes
C (s) |s = 0,

A (s) |s = z0 + (S + 1) |s ,

D (s) |s = z0 (S 1) |s ,
(28)
while the boundary matrix K+ is transformed into

K+;21 e2p + (2K+;11 K+;22 )ep + 2K+;12
K + = ep
2p
p
0
K+;21 e (K+;11 2K+;22 )e 2K+;12

1 + + +
+
=: +
.
=
(29)

0 +
0
+ 1 + + +
It is worth noticing that for a multi-site spin and multi-mode boson generalization of this model
the actions of all consecutive Lax matrices on the pseudo-vacuum have to be in triangular form
for the above procedure to work. Furthermore will we briefly discuss the diagonalization procedure, when the bosonic Lax matrix Eq. (19) is inserted into the transfer matrix, Eq. (16)and
not Eq. (18) as above. A straight forward calculation shows that then the gauge matrix M1 to
the right of the bosonic Lax matrix and all following Lax and boundary matrices must be in
triangular form. As a consequence, K+ instead of K must be triangular.
We now proceed to the result of the algebraic Bethe ansatz described above. The eigenvalues
(; {j }) of the transfer matrix (16) are obtained within the standard formalism [19]

( + S + z0 )( + (S + 1) z0 ) z1
, {j } =
( + z0 + S)( + z0 (S + 1)) + z1
n

2( + )( + )( + + ) ( j 2 ) ( + j 2 )
(2 + )
( j + 2 ) ( + j + 2 )
j =1

( S + z0 )( (S 1) z0 ) + z1 +
+
( + z0 + S)( + z0 (S + 1))
+ z1

n
3
( j + 3
2( + )( + + )
2 ) ( + j + 2 )
. (30)
(2 + )
( j + 2 ) ( + j + 2 )
j =1
The pseudo-vacuum | has eigenvalue 0 () = (; ). The consequence of choosing L(b,2)

as bosonic Lax operator consists in an overall minus sign and z1 z1 . Eigenstates are characterized by the roots {j } of the Bethe equations

(i + (S 12 ) + z0 )(i + (S + 12 ) z0 ) i z1 2 (i + + 2 )(i + 2 )
(i (S + 12 ) + z0 )(i (S 12 ) z0 ) i + z1 + 2 (i + + 2 )(i + 2 )
(i j + )(i + j + )
,
=
(31)
(i j )(i + j )
j =i
292
where the boundary matrix K+ has been rescaled with 1 + + + and + := + / 1 + + + is

the inverse of the effective boundary field. Equations (31) coincide with the Bethe equations for
an open chain with diagonal boundary matrices after a rotation of the spin and bosonic degrees
of freedom. Choosing L(b,2) as the bosonic Lax operator induces the change z1 z1 .
We close this paragraph commenting on the general implications of the restriction to triangular K . Though the boundary conditions do break the su(2) symmetry of the system, there
is a remaining u(1) conserved charge in the system in this case. Hence, the spinboson models
which can be diagonalized within this approach should not be expected to feature counter rotating terms. Nevertheless for non-diagonal triangular K , the resulting model will show some
features of counter rotating terms. We emphasize, however, that the triangularity of K is a constraint necessary only for this method for the diagonalization of the transfer matrix which is not
related to the integrability of the corresponding spinboson model.
4.2. Solution by algebraic contraction of the spinspin transfer matrix
It is well known that bosonic degrees of freedom can be realized as spin operators, in a certain limit. Such a limit constitutes an example of algebraic contraction [42]. In this section we
demonstrate how this idea can be applied to the different stages of QISM method, thus allowing
to obtain integrable spinboson model from certain auxiliary integrable spin model. The approach is pursued here in the case the integrable model is with open boundary conditions, thus
extending the theory for the toroidal boundaries [22].
At the level of the Lax matrices, the spin to boson mapping is achieved by the combination
of two singular transformations, one in the auxiliary and the other in quantum space, to have a
finite result in the limit. Explicitly, the bosonic Lax matrix is obtained as [22]
L(b) () = lim k()L(J ) ( ),
(32)
where the matrix k() is a diagonal matrix

1
k() = diag 1, 2 ,
= + 2 /2,

J = 2 /2,
(33)
2
and the quantum space transformation is defined by {J , 12 J + , J z } {a , a, a a + 2 }. We

have assumed = = in the definition of the Bose Lax matrix, Eq. (18) for convenience.
We apply the Eq. (32) to an auxiliary double-row transfer matrix ta () that in the limit of
is equal to t ()
1

1 (J )
1
ta () = tr K+ ( + ) L(s) ( z0 )
L ( z1 )
k() K ()

k()L(J ) ( z1 )L(s) ( z0 ) .
(34)
The above operator ta () can be conveniently rotated into ta () through a new transformation
Uj (7) acting in the quantum spaces, along the line suggested by the property (6). This simplifies
the effect of k(), acting on the auxiliary space. The transformed ta () reads
ta () = U ta ()U 1

1 (s)
1 (J )
1
L ( z0 )
L ( z1 ) K ()
= tr k()K+ ( + ) k()

L(J ) ( z1 )L(s) ( z0 ) .
(35)
293
Therefore, we can obtain the spectral properties of ta () by the analysis of ta (), whose eigenvalues were previously obtained by means of algebraic Bethe ansatz [32]. In this rephrasing the
K-matrix we need to take into account is

1
K + () = k()K+ () k()

+ + 2 +
= 1
(36)
.
+ +
2
Of course, the eigenvalues of the transfer matrix ta () and ta () are identical and given by the
following expression

( + S + z0 )( + + S z0 )
a () =
(S + + z0 )((S + 1) z0 )

( + J + z1 )( + + J + z1 )
(J + + + z1 )((J + 1) z1 )

n
2( + )( + )( + + ) ( j 2 ) ( + j 2 )
(2 + )
( j + 2 ) ( + j + 2 )
j =1

( S + z0 )( + S z0 )
(S + + z0 )((S + 1) z0 )

2( + )( + + )
( J + z1 )( + J + z1 )
(J + + + z1 )((J + 1) z1 )
(2 + )
n
3
3
( j + ) ( + j + )
2
2
,
(37)
( j + 2 ) ( + j + 2 )
j =1
while the Bethe ansatz equations are given by

(j 2 + S + z0 )(j + 2 + S z0 )
(j 2 S + z0 )(j + 2 S z0 )

(j + 2 + J + z1 )(j + 2 + J z1 )
(j + 2 J + z1 )(j + 2 J z1 )

n
j + 2
j + + 2
(j i + ) (j + i + )
=
,
(j i ) (j + i )
j + 2
j + + 2
(38)
i=1
i=j
, which do not depend on . For the algebraic Bethe ansatz, one of the
where = 1+

two following assumptions on the parameters , and has been made (see [32])
1 + 1 + 1 + 1 + + +
=
, or
(39)
+ 2
1 + 1 + + +
(40)
=
.
+ 2
1 + 1
To obtain the eigenvalues and the Bethe ansatz equation of the spinboson problem, associated to the transfer matrix t () we take the limit of Eqs. (38). This leads to the same
eigenvalue (30) and Bethe ansatz equations (31) as obtained from the direct ansatz. In the limit
294
of the constraints (39), (40) become = 0 and = 0, respectively. This means

that the boundary matrix K () has to be in a triangular (or even diagonal) form already, and
= .
Considering = = 0 in the eigenvalue expression and Bethe ansatz equation above we
can recover the results obtained for the case of diagonal K () K-matrices without use of the
limit process.
As final remark we observe that a definite limit 0 exists, despite the matrices k() are
singular in the limit; such singularity is compensated by the divergence in the impurity (see
Eq. (33)).
5. Integrable Hamiltonians
The transfer matrix t () for rational solutions of the YBE is a polynomial expression in the
spectral parameter. Because the transfer matrix is a commuting family of operators in , the
coefficients of such a polynomial are the integrals of the motion of the theory. In the following
section we exploit this property of the transfer matrix to extract the Hamiltonian. In Section 5.2
we obtain the Hamiltonian following an alternative procedure that is in the spirit of the quasiclassical expansion.
5.1. Integrable models derived from the transfer matrix
We analyze the integrals of the motion generated by the transfer matrix (16) and the bosonic
Lax matrix L(b) = L(b,2) for spin-1/2. We therefore expand it in powers of the spectral parameter
t () =:
6

Hn n
(41)
n=0
where we set = 1.
There are only two independent constants of the motion in this problem and we find that H6 ,
H5 , and H0 are constants and the relations
2H4 H3 = 5 + ,
(42)
H4 H2 + H1 = 3 + .
(43)
With this result for the general case we are now ready for a systematic case study.
At first, we focus on diagonal boundary matrices K . In this case we find up to a constant

H4 = 4 S z n .
(44)
This constant of the motion readily unveils this Hamiltonian to be of the JaynesCummings type
in the RWA because the Hilbert space splits up in two-dimensional (spin-1/2) subspaces with
fixed eigenvalue of H4 . Problems in absence of a conserved number operator in general will
require more sophisticated techniques as employed for the diagonalization of the XYZ chain or
the XXZ chain with non-diagonal boundary fields [34,35].
For diagonal K+ and non-diagonal but triangular K , the expansion (41) leads to (up to a
constant)

1
H4 = 0 S z n + C1 1
(45)
+ 2S z s S + + C2 a + C3 a 2S z n
2s
295
where s is the spin, 0 = 4 , C1 = (2+ + 2z0 1), C2 = 2 (z1 + + ), and C3 = 2 .

As we will show below, the Hilbert space of the non-Hermitian Hamiltonian (45) is not reduced
to two-dimensional subspaces; therefore, it describes a resonant spinboson model (i.e. with
vanishing de-tuning of the spinboson frequencies) containing counter-rotating terms. It can be
treated by the method used above. A careful analysis reveals that the non-Hermiticity of the
Hamiltonian is inherent to this transfer matrix unless the spin degrees of freedom are decoupled
from the boson. We attribute this feature to the structure of the bosonic Lax matrix (18) used
here.
In what follows we restrict ourselves to spin s = 1/2. A vacuum state for the Bethe ansatz on
the Hamiltonian (45) then is |0, 12 , using the notation |n, such that S z |n, = |n, and
n|n,
= n|n, . Defining C2 =: C1 + , C3 = C1 + + , the creation operator B() has the
form

B() = 2 ( + 1) 2 + 2S z a (n ) + n(
n 1)

2 n + 2 ( + 1) a + 2 a n)
S

2 + + S .
(46)
A right-eigenstate |1 , . . . , N := B(1 ) B(N )|0, 12 involves all states |n, with n N
except |N, 12 . It is interesting to notice that for positive integer values of the inverse boundary
field the Hilbert space accessible from the vacuum |0, 12 is restricted to

H N+ := |n, ; n .
(47)
Without loss of generality we set g1 = 1 and + = + = 0. The latter corresponds to a diagonal boundary matrix K+ () = + 1 + z ; any non-diagonal K+ can be diagonalized by a
gauge transformation [27,32] and proper r0 , s0 , r0 , s0 . For convenience, we choose = = 1
and rename g2 =: g. The most general Hamiltonian generated in this way is equivalent to
H =
g1 H4 + g2 H1
=: Hs + Hb + Hsb + Hsbb
4g1
(48)
via suitable canonical transformations on the spin and bosonic degrees of freedom.
Our goal will be to have the spin and boson Hamiltonian together with major part of the spin
boson interaction Hermitian. To this end, we set = + ( g1 + + z1 ), which makes the occurring
0 2+ 1
terms aS
z and a S z Hermitian and the choice = 1 z1 2z

2z0 +2+ 1 realizes the same for aS
and a S + . z1 = 1/2 eliminates a term nS
+ and the pure spin Hamiltonian becomes Hermitian
if
g=4
4
2+ 1
+ 16
.
+
2z0 6+ 1
(49)
The terms a and a in the pure boson part of the Hamiltonian are obtained Hermitian
if

1 g(g + 2)+2 + g 2 (g + 1)+4
1
z0 =
(50)
2
g(g+2 1)
and it turns out that both conditions (49) and (50) are compatible and fixes a one-dimensional
manifold in + z0 space (see Fig. 1).
296
Fig. 1. Left: Solutions of the non-linear equation in + and g emerging from inserting Eq. (50) into (49). Right: The
corresponding values for z0 .
The resulting Hamiltonian is as follows

1 4+2
1 + 2+
+ (2+ 1)
Hs = 4
4
8
Sz
+
2z0 + 2+ 1
2z0 6+ 1
(+ 1)((2z0 1)2 4+2 ) x
(51)
S ,
+ (2z0 6+ + 1)
2

4 2
+ (g+ + 2)
2 +
a
+
a
a n,
Hb = g 2 2 + n g
(52)
+
g
g+ 1
g+2 1
g+2 1

z + 2 (g+ + 2) a + a S z + 2g(2z0 2+ 1) aS
+ a S + ,
Hsb = 8g+ nS
(53)

g+ + 2 2 +
Hsbb = g (2z0 + 2+ 1)
a S (2z0 2+ 1) a 2 S
2g+

+ g(2z0 + 2+ 1) n 2 S + + 2a nS
(54)
+ 4g + a nS
z.
+ 2
Besides the Hermitian parts Hs and Hsb , there are many non-Hermitian terms, in particular
non-linear couplings as in Hsbb . It is worth focusing on the non-Hermitian term in Hb proportional to a n.
This term has an immediate physical interpretation: it describes a leakage
of bosons/photons from the cavity, where the escape rate is proportional to the number of
bosons/photons (i.e. the intensity of the cavity boson field).
5.2. Integrable models derived from the quasi-classical limit of the transfer matrix
Another way to obtain an integrable Hamiltonian is to take the so-called quasi-classical limit
of the transfer matrix [20,21,29]. It consists in a series expansion in the quantum parameter
of the transfer matrix around = 0: t() = (0) + (1) () + 2 (2) () + with the aim of
creating a commuting family of quasi-classical transfer matrices (k) (). This procedure may be
particularly useful for extracting simple though non-locally interacting Hamiltonians out of the
transfer matrix. Examples are the Gaudin magnets and corresponding BCS-like models.
There is a wide freedom of introducing an -dependence to the boundary matrix parameters
in the theory, without destroying its integrability. For the case of integrable spin chains with diagonal boundaries [29] this procedure was a valid tool to define integrable one-parameter extension
297
of Gaudin models in non-uniform magnetic fields [25]; such models describe metallic grains
with pairing and magnetic interaction [24].
In the present case the transfer matrix is the finite sum
t() = k (k) + + (0) + (1) () + 2 (2) () + + m (m) ()
(55)
where k and m are integers. Expanding the commutator relation [t(), t( )] = 0 in , we obtain
2m

t(), t( ) =
l Cl (, ) = 0,
l=2k
which implies Cl (, ) = 0 for all l. The first relevant terms are

C2k (, ) = (k) (), (k) ( ) ,

C2k+1 (, ) = (k) (), (k+1) ( ) + (k+1) (), (k) ( ) ,

C2k+2 (, ) = (k) (), (k+2) ( ) + (k+2) (), (k) ( )

+ (k+1) (), (k+1) ( ) .
(56)
From the expressions above, one finds that the first (n) () which is not a C-number (times
the identity) gives rise to a family of commuting operators. Generically, the lowest order (k) is a C-number. Therefore the first class of integrable models were generated by
[ (k+1) (), (k+1) ( )] = 0. In the presence of boundary matrices, these are typically nontrivial operators but representing non-interacting Hamiltonians. The task is then to tune the free
parameters such that (k+1) () is also a C-number, and that the lowest non-trivial order in ,
e.g. (k+2) (), be an interesting Hamiltonian.
Here we allow , , to be generic functions of whose form can be fixed to modify to our convenience the -expansion of the transfer matrix, and ultimately the Hamiltonian.
Specifically, we write the boundary parameters as
(1) 1
(0)
(1)
x = x
+ x + x
(57)
(1)
(1)
(1)
with x = {, , }, and set = = + = 0. Then the
the expansion, proportional to 2 , already contains non-trivial
lowest non-vanishing term in

operators; they can be made
(0)
=
=
0.
Analogously,
the
next
term
in the expansion can be
vanishing by the choice (0)
+
(0)
(0)
made proportional to the identity by + = + = 0 and z1 = 0. The next non-trivial term in the
-expansion (0) () can be used as Hamiltonian
.
H=
1
2

(0) () = 40 a a + 2 a + a + z S z + 4 z0 Sx

+ GS z a a 4 z0 aS + a S +

2
+ 2 S + S + S S + 2 S z + C1
(1)
(1)
(58)
(1)
where 0 = (2 z02 ), = + and z = 42 2 2 (1 2+ ), G = 4 2 (1 )

1 )2 + 2 0 ] 2( z )( 2 0 )}.
and C = z1 {0 [2(2 + 0
0
For S = 1/2 the Hamiltonian simplifies to

H1/2 = 0 a a + z S z + GS z a a + GS z a + a g S + a + S a
(59)
298
where the bosonic operators have been displaced a a + /2, z = z + 2 G and a term
proportional to the identity was omitted. The Hamiltonian (59) display the rotating and counterrotating simultaneously. This can be evidenced by a rotation R of the spin

RH1/2 R 1 = 0 a a + z cos(2 )S z sin(2 )S x + G cos(2 )S z sin(2 )S x a a

cos(2 ) + 1 +
cos(2 ) 1 +
g
(60)
S a + Sa g
S a + S a ,
2
2
.
where R = exp[ (S + S )] and tan = G/g has been chosen to eliminate the S z (a + a )
term.
6. Conclusions
By the Quantum Inverse scattering method we have constructed integrable models with
twisted and open boundary conditions which contain rotating as well as counter rotating terms.
Twisted boundary conditions lead to Hamiltonians where solely rotating (or counter-rotating)
terms appear.
In case of open boundaries, we diagonalize the models by algebraic Bethe ansatz when at
least one of the boundary matrices has triangular form. To this end, we extended the procedure
for spin chains to models which are derived from a particular bosonic Lax matrix besides the
standard spin Lax matrix. We further analyzed generic boundary matrices in the sense that they
cannot be brought both in triangular form at the same time. For this choice of boundary fields,

the symmetry leading to conservation of the number operator S z + j aj aj is broken.
Taking the full transfer matrix the integrable Hamiltonians are non-Hermitian. However, a
parameter choice is found that shifts the non-Hermiticity to non-linear terms in the spinboson
interaction and the bosonic part of the Hamiltonian. In the latter, the non-Hermiticity can be
interpreted as a density dependent leakage of photons out of the cavity. Different parameter
choices could be analyzed in order that non-Hermitian terms will appear elsewhere, e.g. in the
spin part of the Hamiltonian. It is worth studying what type of non-Hermiticity are most realistic
in order to tailor the parameters closest to this situation. The diagonalization of the resulting
Hamiltonians together with a discussion of the effects of the undesired non-Hermitian parts is
left to forthcoming work.
Interestingly enough, by a procedure that is close in spirit to the quasi-classical limit, also
Hermitian Hamiltonians can be extracted out of the transfer matrix. We have given an explicit
example, which contains both rotating and counter-rotating terms. The evidence we have for this
is that any eigenstate (i.e. coherent states) of the bosonic part is spread out all over the bosonic
Hilbert space by means of the interaction term.
Though integrable counter-rotating spinboson models have been formulated in this work,
it must be emphasized that their diagonalization with e.g. the algebraic Bethe ansatz is still an
open problem. It could be interesting to study general open boundary conditions still for the
XXX model but based on the HolsteinPrimakoff representation of su(2). The advantage of
such an approach is that the Hamiltonian is constructed in a straight forward manner from the
existing solution of the open XXX spin chain Hamiltonian and hence also the resulting spin
boson Hamiltonian is Hermitian by construction. The challenge again lies in the diagonalization
of the transfer matrix. Finally, the analysis of spinboson models fanning out from a bulk with
XYZ-symmetry (parametrized by elliptic functions) constitute an interesting challenge for future
investigation.
299
Acknowledgements
We thank A. Kundu for fruitful discussions; G.A.P. Ribeiro thanks FAPESP (Fundao de
Amparo Pesquisa do Estado de So Paulo) for financial support and the DMFCI for the hospitality.
References
[1] C. Cohen-Tannoudji, J. Dupont-Roc, G. Grynberg, AtomPhoton Interactions, Wiley, New York, 1992.
[2] E.T. Jaynes, F.W. Cummings, Proc. IEEE 51 (1963) 89.
[3] M. Tavis, F.W. Cummings, Phys. Rev. 170 (1969) 379;
M. Tavis, F.W. Cummings, Phys. Rev. 188 (1969) 692.
[4] U. Weiss, Dissipative Quantum Mechanics, World Scientific, Singapore, 1999.
[5] M.H. Devoret, A. Wallraff, J.M. Martinis, cond-mat/0411174.
[6] M. Paternostro, et al., Phys. Rev. B 69 (2004) 214502;
F. Plastina, G. Falci, Phys. Rev. B 67 (2003) 224514.
[7] J.I. Cirac, et al., Phys. Rev. A 46 (1992) 2668.
[8] R.J. Hughes, et al., Fortschr. Phys. 46 (1998) 329.
[9] J.E. Mooij, et al., Science 285 (1999) 1036.
[10] S. Datta, B. Das, Appl. Phys. Lett. 56 (1990) 665;
L.W. Molenkamp, G. Schmidt, G.E.W. Bauer, Phys. Rev. B 64 (2001) 121202.
[11] J. Schliemann, J.C. Egues, D. Loss, Phys. Rev. B 67 (2003) 085302.
[12] H.-A. Engel, E. Rashba, B. Halperin, in: H. Kronmller, S. Parkin (Eds.), Handbook of Magnetism and Advanced
Materials, vol. 5 cond-mat/0603306.
[13] C. Emary, T. Brandes, Phys. Rev. E 67 (2003) 066203.
[14] K. Hepp, E. Lieb, Ann. Phys. 76 (1973) 360.
[15] A. Wallraff, et al., cond-mat/0703632.
[16] A.G. Maugeri, G. Falci, in preparation.
[17] V.E. Korepin, N.M. Bogoliubov, A.G. Izergin, Quantum Inverse Scattering Method and Correlation Functions,
Cambridge Univ. Press, Cambridge, 1993.
[18] H.J. de Vega, Nucl. Phys. B 240 (1984) 495.
[19] E.K. Sklyanin, J. Phys. A: Math. Gen. 21 (1988) 2375.
[20] E.K. Sklyanin, J. Sov. Math. 47 (1989) 2473.
[21] K. Hikami, P.P. Kulish, M. Wadati, J. Phys. Soc. Jpn. 61 (1992) 3071.
[22] L. Amico, H. Hikami, Eur. Phys. J. B 43 (2005) 387.
[23] E. Sklyanin, J. Sov. Math. 47 (1989) 2473.
[24] L. Amico, A. Di Lorenzo, A. Osterloh, Nucl. Phys. B 614 (2001) 449.
[25] A. Di Lorenzo, et al., Nucl. Phys. B 644 (2002) 409.
[26] J. von Delft, R. Poghossian, Phys. Rev. B 66 (2002) 134502.
[27] G.A.P. Ribeiro, M.J. Martins, Nucl. Phys. B 705 (2005) 521.
[28] H. Johannesson, Nucl. Phys. B 270 (1986) 235.
[29] K. Hikami, J. Phys. A 28 (1995) 4997.
[30] I. Cherednik, Theor. Math. Phys. 61 (1984) 977.
[31] H.J. de Vega, A. Gonzles-Ruiz, J. Phys. A 27 (2004) 6129.
[32] C.S. Melo, G.A.P. Ribeiro, M.J. Martins, Nucl. Phys. B 711 (2005) 565.
[33] W. Galleas, M.J. Martins, Phys. Lett. A 335 (2005) 167.
[34] R. Murgan, R.I. Nepomechie, C. Shi, hep-th/0605223.
[35] J. Cao, H.-Q. Lin, K.-J. Shi, Y. Wang, Nucl. Phys. B 663 (2003) 487.
[36] V. Buek, Phys. Rev. A 39 (1989) 3196.
[37] N.M. Bogoliubov, R.K. Bullough, J. Timonen, J. Phys. A: Math. Gen. 29 (1996) 6305.
[38] B. Jurco, J. Math. Phys. 30 (1989) 1739.
300
[39] A. Rybin, et al., J. Phys. A: Math. Gen. 31 (1998) 4705.

[40] A. Kundu, Phys. Rev. Lett. 82 (1999) 3936;
A. Kundu, quant-ph/0307102.
[41] W.-L. Yang, Y.-Z. Zhang, M. Gould, Nucl. Phys. B 698 (2004) 503;
W.-L. Yang, Y.-Z. Zhang, JHEP 0704 (2007) 044.
[42] R. Gilmore, Lie Groups, Lie Algebras, and Some of Their Applications, Wiley, New York, 1974.

Nucl - Phys.B v.787

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Nucl - Phys.B v.787

Diunggah oleh

Hak Cipta:

Format Tersedia

Nuclear Physics B 787 (2007) 154

Squark and gaugino hadroproduction and decays

53 Avenue des Martyrs, F-38026 Grenoble, France

E-mail address: klasen@lpsc.in2p3.fr (M. Klasen).

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

by (arbitrary) non-diagonal entries ij with i, j = L, R in the squared squark mass matrices

where the diagonal elements are given by

for up-type squarks,

for down-type squarks,

while the well-known squark helicity mixing is generated by the elements

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

tional 6 6 matrices R u and R d with

In the limit of vanishing off-diagonal parameters ij , the matrices R q become flavour-diagonal,

qi (p1 )qj (p2 ),

j (p1 )qi (p2 ),

g and expressed in terms

d 1,1 + d 1,1 + d 1,1 + d 1,1

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

3.1. Generalized strong and electroweak couplings in NMFV SUSY

the standard notation, the W i j , Z i j , and Z i0 j0 interaction vertices

OijL = Ni4 Vj2 + Ni2 Vj1 and OijR = Ni3

Oij L = Vi1 Vj1 Vi2 Vj2 + ij xW and Oij R = Ui1

{Lqq Z , Rqq Z } = 2Tq3 2eq xW qq ,

{Lqi qj Z , Rqi qj Z } = 2Tq3 2eq xW

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

mdk cW Ni3 Rjd(k+3)

mdk cW Ni3 Rjdk

muk cW Ni4 Rju(k+3)

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

thanks to the form factors

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

involves many different form factors,

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

Re Cqi q g Cqj q g ut m2qi m2qj ,

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

3.3. Squark pair production

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

involves the form factors

ut m2d m2u j (1 mn ) + m 0 m 0 smn ,

ut m2d m2u j (mn 1) + m m 0 smn ,

ut m2d m2u j (mn 1) + m mg smn ,

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

[GU ]21 [GT U]21

depends therefore on the form factors

ut m2qi m2qj (1 mn ) + m 0 m 0 smn ,

ut m2qi m2qj (1 mn ) + m 0 m 0 smn ,

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

ut m2qi m2qj (1 mn ) + m 0 mg smn ,

2(ut m2qi m2j ) 

(1 ha )(1 hb )|Lqi q j |2 + (1 + ha )(1 + hb )|Rqi q j |2

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

Fig. 6. Tree-level Feynman diagrams for the production of gaugino pairs.

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

while those of the latter are given by

|Lqi qk Z + Rqi qk Z |2 3/2 m2qi , m2Z , m2qk ,

|Lqi qk W |2 3/2 m2qi , m2W , m2q .

The usual Kllen function is

In cMFV, our results agree with those of Ref. [31].

G. Bozzi et al. / Nuclear Physics B 787 (2007) 154

2(ut m2qi m2j )